Power signal type determination method and device and electronic equipment

Abstract

The invention discloses a power signal type determination method and device and electronic equipment. The method comprises the following steps: acquiring an original signal sequence; determining an amplitude feature sequence corresponding to the original signal sequence; according to a time point sequence, a first characteristic parameter corresponding to a first amplitude characteristic value is determined according to the first amplitude characteristic value corresponding to a first time point, and a second characteristic parameter corresponding to a second amplitude characteristic value is determined according to the first characteristic parameter and a second amplitude characteristic value corresponding to a second time point; obtaining a target characteristic parameter until a plurality of amplitude characteristic values in the amplitude characteristic sequence are processed; and determining a target signal type corresponding to the original signal sequence according to the target characteristic parameter. According to the method and the device, the technical problem that the type of the electric power signal is determined inaccurately due to the fact that the electric power signal is interfered by various noises due to diversification and complexity of an electromagnetic interference source when the type of the electric power signal is determined in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular, to a method, apparatus, and electronic device for determining the type of power signal. Background Art

[0002] In the related art, when determining the type of the collected power signal, due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in the technical problem of inaccurate determination of the power signal type.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of the present invention provide a method, apparatus, and electronic device for determining the type of power signal, so as to at least solve the technical problem in the related art that when determining the type of the power signal, due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in inaccurate determination of the power signal type.

[0005] According to an aspect of an embodiment of the present invention, a method for determining the type of power signal is provided, including: obtaining an original signal sequence; determining an amplitude feature sequence corresponding to the original signal sequence, where the amplitude feature sequence includes amplitude feature values corresponding to multiple time points arranged in chronological order; according to the time point order, based on the first amplitude feature value corresponding to the first time point, determining a first feature parameter corresponding to the first amplitude feature value, and based on the first feature parameter and the second amplitude feature value corresponding to the second time point, determining a second feature parameter corresponding to the second amplitude feature value, until all the amplitude feature values in the amplitude feature sequence are processed, to obtain a target feature parameter; and determining a target signal type corresponding to the original signal sequence according to the target feature parameter.

[0006] Optionally, the determining the amplitude feature sequence corresponding to the original signal sequence includes: determining a signal conversion method corresponding to the original signal sequence; according to the signal conversion method, determining multiple signal comparison sequences corresponding to the original signal sequence, where the multiple signal comparison sequences are sequences used to compare with the original signal sequence; and determining the amplitude feature sequence corresponding to the original signal sequence according to the multiple signal comparison sequences and the original signal sequence.

[0007] Optionally, determining a plurality of signal comparison sequences corresponding to the original signal sequence according to the signal conversion method includes: determining a plurality of scale parameters corresponding to the signal conversion method and a plurality of time parameters corresponding to the corresponding scale parameters; determining an initial comparison sequence corresponding to each of the plurality of scale parameters according to the signal conversion method; and determining a plurality of signal comparison sequences corresponding to the original signal sequence according to the initial comparison sequences corresponding to the plurality of scale parameters and the plurality of time parameters corresponding to the corresponding scale parameters.

[0008] Optionally, determining the amplitude feature sequence corresponding to the original signal sequence includes: determining the amplitude range parameter corresponding to the original signal sequence; determining the amplitude filtering sequence corresponding to the original signal sequence according to the amplitude range parameter corresponding to the original signal sequence; and determining the amplitude feature sequence corresponding to the original signal sequence according to the amplitude filtering sequence and the original signal sequence.

[0009] Optionally, determining the first feature parameter corresponding to the first amplitude feature value according to the first amplitude feature value corresponding to the first time point includes: determining the initial time point corresponding to the amplitude feature sequence and the initial feature parameter corresponding to the initial time point; determining the first screening index corresponding to the first amplitude feature value according to the initial feature parameter and the first amplitude feature value corresponding to the first time point; and determining the first feature parameter corresponding to the first amplitude feature value according to the first screening index.

[0010] Optionally, determining the first feature parameter corresponding to the first amplitude feature value according to the first screening index includes: when the first screening index includes a first retention index, a first update index, and a first weight index, determining the sequence feature parameter corresponding to the first amplitude feature value according to the first retention index and the first update index, where the first retention index is used to represent the retention degree of the initial feature parameter, the first update index is used to represent the update degree of the first amplitude feature value to the initial feature parameter, and the first weight index is used to represent the importance of the first amplitude feature value for characterizing the amplitude feature sequence; and determining the first feature parameter corresponding to the first amplitude feature value according to the first weight index and the sequence feature parameter corresponding to the first amplitude feature value.

[0011] Optionally, determining the target signal type corresponding to the original signal sequence according to the target feature parameter includes: determining a linear transformation parameter and a threshold parameter; determining a linear feature parameter corresponding to the original signal sequence according to the linear transformation parameter and the target feature parameter; and determining the target signal type corresponding to the original signal sequence according to the linear feature parameter and the threshold parameter.

[0012] According to one aspect of an embodiment of the present invention, there is provided a power signal type determination device, including: an acquisition module configured to acquire an original signal sequence; a first determination module configured to determine an amplitude feature sequence corresponding to the original signal sequence, where the amplitude feature sequence includes a plurality of amplitude feature values arranged in chronological order; a second determination module configured to, in chronological order, determine a first feature parameter corresponding to a first amplitude feature value corresponding to a first time point according to the first amplitude feature value, and determine a second feature parameter corresponding to a second amplitude feature value corresponding to a second time point according to the first feature parameter and the second amplitude feature value, until a plurality of amplitude feature values in the amplitude feature sequence are processed to obtain a target feature parameter; and a third determination module configured to determine a target signal type corresponding to the original signal sequence according to the target feature parameter.

[0013] According to one aspect of an embodiment of the present invention, there is provided an electronic device, including: a processor; a memory configured to store instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the power signal type determination method according to any one of the above.

[0014] According to one aspect of an embodiment of the present invention, there is provided a computer-readable storage medium, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, enabling the electronic device to execute the power signal type determination method according to any one of the above.

[0015] In an embodiment of the present invention, an original signal sequence is obtained; an amplitude feature sequence corresponding to the original signal sequence is determined, where the amplitude feature sequence includes amplitude feature values corresponding to multiple time points arranged in chronological order; in the order of time points, according to the first amplitude feature value corresponding to the first time point, a first feature parameter corresponding to the first amplitude feature value is determined, and according to the first feature parameter and the second amplitude feature value corresponding to the second time point, a second feature parameter corresponding to the second amplitude feature value is determined, until all the amplitude feature values in the amplitude feature sequence are processed, and a target feature parameter is obtained; according to the target feature parameter, a target signal type corresponding to the original signal sequence is determined. The original signal sequence can reflect the working state of the device to be detected, providing a data basis for determining the signal type corresponding to the original signal subsequently. By extracting the amplitude feature information related to time from the original signal sequence, it is helpful to more intuitively reflect the change of signal intensity, and further helpful for accurately judging the type of power signal subsequently. By analyzing the features of multiple amplitude feature values in the amplitude feature sequence in the order of time points, the changes of the power signal at each time point can be effectively captured and analyzed, and the overall features of the amplitude feature sequence can be accurately identified, so that the features of the power signal can be accurately captured. Using the obtained target feature parameter of the power signal for classification helps to accurately identify the specific type of the power signal, thus ensuring the accuracy and reliability of the determination of the power signal type in a complex environment such as the existence of high-frequency pulse interference, and further solving the technical problem in the related art that when determining the type of the power signal, due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in inaccurate determination of the power signal type. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0017] Figure 1 is a flowchart of the method for determining the type of power signal according to the embodiment of the present invention;

[0018] Figure 2 is a structural block diagram of the device for determining the type of power signal according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0020] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings 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 under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] First, some nouns or terms that appear during the description of the embodiments of the present application are applicable to the following explanations:

[0022] SureShrink: SureShrink is a denoising method based on wavelet transform. In wavelet denoising, the SureShrink strategy uses soft thresholding or hard thresholding to eliminate noise in wavelet coefficients. It first decomposes the signal into different frequency bands through wavelet transform, and then applies a threshold operation to the wavelet coefficients to compress or set to zero those coefficients that are considered to contain noise, thereby achieving the denoising effect. The key of SureShrink lies in its ability to automatically select the optimal threshold, avoiding the subjectivity and uncertainty of manually setting the threshold.

[0023] Adam optimization algorithm: Adam (Adaptive Moment Estimation) is an optimization algorithm with an adaptive learning rate, commonly used for training deep learning models such as neural networks. Adam combines the advantages of the AdaGrad and RMSprop algorithms, calculates the first-order moment estimate and second-order moment estimate of the gradient, and performs bias correction to dynamically adjust the learning rate of each parameter to improve the training speed and effect.

[0024] Dropout Layer: Dropout is a regularization technique used in deep learning to prevent model overfitting. During training, the Dropout layer "drops out" (i.e., sets to 0) neurons in the neural network and all their participating connections with a certain probability. Dropout is usually used before fully connected layers, and its dropout rate is a hyperparameter that needs to be adjusted according to the specific task.

[0025] LSTM Layer: LSTM (Long Short-Term Memory) is a special type of Recurrent Neural Network (RNN) layer designed to process and predict sequential data. Memory cells and gating mechanisms (input gate, output gate, and forget gate) are introduced in LSTM.

[0026] Fully Connected Layer: The fully connected layer is a common layer in neural networks, where each neuron in the layer is connected to all neurons in the previous layer. In a deep learning model, the fully connected layer is usually located at the end of the model and is used to summarize the previous feature representations and map them to the final output classification or regression results. The fully connected layer performs a linear transformation through a weight matrix and a bias vector, and then a non-linear transformation through an activation function to achieve classification or regression prediction of the input features.

[0027] Embodiment 1

[0028] According to an embodiment of the present invention, an embodiment of a method for determining the type of power signal is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] Figure 1 is a flowchart of the method for determining the type of power signal according to an embodiment of the present invention, as Figure 1 shown, the method includes the following steps:

[0030] S102, obtain the original signal sequence;

[0031] In step S102 provided in the present application, the original signal sequence is obtained.

[0032] Among them, the original signal sequence is involved. The original signal sequence is a signal sequence obtained by collecting and used for signal type judgment. The original signal sequence includes a plurality of original signal values arranged in chronological order. The original signal sequence can be the original power signal data collected and recorded from the cable operating environment.

[0033] The working state of the device to be detected can be reflected by the original signal sequence. For example, when detecting partial discharge in a cable, by analyzing the original signal sequence of the cable, the partial discharge situation of the cable can be understood. Therefore, by obtaining the original signal sequence, a data basis can be provided for subsequent determination of the signal type corresponding to the original signal.

[0034] S104. Determine the amplitude feature sequence corresponding to the original signal sequence, where the amplitude feature sequence includes amplitude feature values corresponding to multiple time points arranged in chronological order.

[0035] In step S104 provided in this application, the amplitude feature sequence corresponding to the original signal sequence is determined.

[0036] Among them, the amplitude feature sequence is involved. The amplitude feature sequence is obtained by converting the signal values in the original signal sequence and can reflect the amplitude features of the original signal in the time domain and frequency domain.

[0037] Among them, multiple time points are involved. The multiple time points are the respective time nodes for recording data during the signal acquisition process. Each time point corresponds to a sample point in the amplitude feature sequence and reflects the amplitude feature of the signal at that time point.

[0038] Among them, the amplitude feature value is involved. The amplitude feature value is used to represent the amplitude or intensity of the signal at the corresponding time point.

[0039] By extracting the time-related amplitude feature information from the original signal sequence to form an amplitude feature sequence representing the signal intensity, it is helpful to more intuitively reflect the change of the signal intensity, and further helpful for subsequent accurate judgment of the type of power signal.

[0040] S106. According to the time point order, based on the first amplitude feature value corresponding to the first time point, determine the first feature parameter corresponding to the first amplitude feature value, and based on the first feature parameter and the second amplitude feature value corresponding to the second time point, determine the second feature parameter corresponding to the second amplitude feature value, until all the amplitude feature values in the amplitude feature sequence are processed to obtain the target feature parameter.

[0041] In step S106 provided in this application, the target feature parameter is obtained.

[0042] Among them, the first time point is involved. The first time point is the first time node when the signal starts to be collected and represents the starting point of data recording.

[0043] Among them, the first amplitude feature value is involved. The first amplitude feature value is the value corresponding to the first time point in the amplitude feature sequence. The first amplitude feature value reflects the amplitude feature of the signal at the first time point.

[0044] Among them, a first characteristic parameter is involved, and the first characteristic parameter is a characteristic used to represent a first amplitude eigenvalue.

[0045] Among them, a second time point is involved, and the second time point is the next time node after the first time node in the amplitude feature sequence arranged in the order of the time series.

[0046] Among them, a second amplitude eigenvalue is involved, and the second amplitude eigenvalue is the value corresponding to the amplitude feature sequence at the second time point. The second amplitude eigenvalue reflects the amplitude feature of the signal at the second time point.

[0047] Among them, a second characteristic parameter is involved, and the second characteristic parameter is a comprehensive characteristic used to represent the second amplitude eigenvalue and the amplitude eigenvalues between the second time points.

[0048] Among them, a target characteristic parameter is involved, and the target characteristic parameter is a comprehensive characteristic used to represent the last amplitude eigenvalue in the amplitude feature sequence and the amplitude eigenvalues corresponding to all previous time points, that is, the sequence characteristic of the amplitude feature sequence.

[0049] By sequentially analyzing the characteristics of multiple amplitude eigenvalues in the amplitude feature sequence in the order of time points, it is possible to effectively capture and analyze the changes of the power signal at each time point, and accurately identify the overall characteristics of the amplitude feature sequence, so as to accurately capture the characteristics of the power signal, which helps to distinguish the types of power signals subsequently, and further ensures the accurate judgment of the power signal type.

[0050] S108. According to the target characteristic parameter, determine the target signal type corresponding to the original signal sequence.

[0051] In step S108 provided in the present application, the target signal type corresponding to the original signal sequence is determined.

[0052] Among them, a target signal type is involved, and the target signal type is a category corresponding to the power signal determined after feature extraction and classification of the amplitude feature sequence of the power signal. The target signal type can be specific types such as partial discharge signals, high-frequency pulse interference signals, periodic narrowband interference signals, random white noise, etc. For example, if after processing and analysis, the power signal is determined to be a signal generated by partial discharge, then the "target signal type" is a partial discharge signal.

[0053] Using the extracted target characteristic parameter of the power signal for classification helps to accurately identify the specific type of the power signal, thus ensuring the accuracy and reliability of the determination of the power signal type in a complex environment such as the presence of high-frequency pulse interference.

[0054] Through the above steps S102 - S108, an original signal sequence is obtained; an amplitude feature sequence corresponding to the original signal sequence is determined, where the amplitude feature sequence includes amplitude feature values corresponding to multiple time points arranged in chronological order; in chronological order of time points, based on the first amplitude feature value corresponding to the first time point, a first feature parameter corresponding to the first amplitude feature value is determined, and based on the first feature parameter and the second amplitude feature value corresponding to the second time point, a second feature parameter corresponding to the second amplitude feature value is determined, until all the amplitude feature values in the amplitude feature sequence are processed, obtaining target feature parameters; based on the target feature parameters, the target signal type corresponding to the original signal sequence is determined. The original signal sequence can reflect the working state of the device to be detected, providing a data basis for subsequent determination of the signal type corresponding to the original signal. By extracting time - related amplitude feature information from the original signal sequence, it is helpful to more intuitively reflect the change in signal intensity, and further helps to accurately judge the type of power signal in the subsequent process. By sequentially analyzing the features of multiple amplitude feature values in the amplitude feature sequence in chronological order of time points, it is possible to effectively capture and analyze the changes in the power signal at each time point, and accurately identify the overall features of the amplitude feature sequence, thus accurately capturing the features of the power signal. Using the obtained target feature parameters of the power signal for classification helps to accurately identify the specific type of the power signal, ensuring the accuracy and reliability of the determination of the power signal type in a complex environment such as the presence of high - frequency pulse interference. Furthermore, it solves the technical problem in the related art that when determining the type of power signal, due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in inaccurate determination of the power signal type.

[0055] As an alternative embodiment, determining the amplitude feature sequence corresponding to the original signal sequence includes: determining the signal conversion method corresponding to the original signal sequence; based on the signal conversion method, determining multiple signal comparison sequences corresponding to the original signal sequence, where the multiple signal comparison sequences are sequences used for comparison with the original signal sequence; based on the multiple signal comparison sequences and the original signal sequence, determining the amplitude feature sequence corresponding to the original signal sequence.

[0056] In this embodiment, the specific steps for determining the amplitude feature sequence corresponding to the original signal sequence are described.

[0057] Among them, a signal conversion method is involved. This signal conversion method is a method of converting an original signal into another form or analyzing the signal in different domains (such as the time domain and the frequency domain). For example, the fast Fourier transform (FFT) can be used to convert a time-domain signal into a frequency-domain signal to analyze the spectral components of the signal; or the wavelet transform can be used to decompose the signal into wavelet coefficients of different scales to identify the local change characteristics of the signal.

[0058] Among them, a signal alignment sequence is involved. This signal alignment sequence is determined based on the signal conversion method and is used to compare and analyze with the original signal sequence to identify a specific pattern or feature in the signal. For example, a signal sequence obtained based on the Daubechies 4th-order wavelet basis function (i.e., the db4 wavelet basis function).

[0059] In the steps involved in this embodiment, first, based on the nature of the power signal and the expected analysis target, an appropriate signal conversion method needs to be determined. For example, the wavelet transform can be used for time-frequency domain signal analysis, while the Fourier transform is more suitable for spectral analysis. Then, according to the selected signal conversion method, one or more groups of signal alignment sequences are generated. These signal alignment sequences are used to compare with the original signal sequence to identify specific features or patterns in the power signal. Finally, through comprehensive comparison and analysis of multiple signal alignment sequences and the original signal sequence, an amplitude feature sequence is obtained.

[0060] Determining the signal conversion method corresponding to the original signal sequence. The function or purpose of this step is to select the most suitable analysis method according to the characteristics of the signal, so as to extract the deep features of the signal. For example, using the wavelet transform can more effectively capture the local changes of the signal and is suitable for non-stationary signal analysis; while the fast Fourier transform (FFT) is good at analyzing the spectral components of the signal and helps to identify the periodic features in the signal. Selecting an appropriate signal conversion method can ensure the accuracy and pertinence of subsequent analysis, and reduce information loss or noise amplification caused by improper signal conversion. This is the basis for determining the amplitude feature sequence.

[0061] By determining multiple signal alignment sequences, diverse perspectives of the signal in different domains or scales are provided, which can help identify the subtle differences and interference patterns in the power signal and provide a multi-dimensional analysis basis for the determination of the amplitude feature sequence. By comparing the original signal sequence with the signal alignment sequences, the effective information in the power signal can be highlighted, and at the same time, the interference components can be suppressed or removed, thus helping to identify and quantify the amplitude features of the power signal, which is helpful for subsequent power signal classification. It can highlight the effective information in the signal and suppress or remove the interference components at the same time.

[0062] As an alternative embodiment, according to the signal conversion method, multiple signal comparison sequences corresponding to the original signal sequence are determined, including: determining multiple scale parameters corresponding to the signal conversion method, and multiple time parameters corresponding to the corresponding scale parameters; according to the signal conversion method, determining initial comparison sequences corresponding to the multiple scale parameters respectively; and determining multiple signal comparison sequences corresponding to the original signal sequence according to the initial comparison sequences corresponding to the multiple scale parameters respectively and the multiple time parameters corresponding to the corresponding scale parameters.

[0063] In this embodiment, the specific steps of determining multiple signal comparison sequences corresponding to the original signal sequence according to the signal conversion method are described.

[0064] Among them, multiple scale parameters are involved. These multiple scale parameters represent different transformation scales set during the signal conversion analysis, and are used to capture the power signal characteristics of the original signal sequence at different scales. For example, when using wavelet transform, a series of scale parameters such as 1, 2, 4, 8, etc. can be selected to capture the high-frequency, medium-frequency, and low-frequency characteristics of the signal respectively. Each scale parameter corresponds to a decomposition level in the wavelet transform. The higher the level, the larger the scale, and the lower-frequency the signal characteristics captured; conversely, the higher-frequency the characteristics captured.

[0065] Among them, multiple time parameters are involved. These multiple time parameters are parameters used to locate the time points or time windows of the power signal during the signal conversion analysis, and are used to control the transformation (such as translation) of the initial comparison sequences at different scales in the time dimension.

[0066] Among them, initial comparison sequences are involved. These initial comparison sequences are a series of comparison sequences generated according to the signal conversion method under the scale parameters and used to analyze the original signal sequence.

[0067] In the steps involved in this embodiment, first, multiple scale parameters corresponding to the signal conversion method, and multiple time parameters corresponding to the corresponding scale parameters are determined. Then, according to the signal conversion method, the initial comparison sequences corresponding to the multiple scale parameters are determined respectively. Finally, according to the initial comparison sequences corresponding to the multiple scale parameters respectively and the multiple time parameters corresponding to the corresponding scale parameters, multiple signal comparison sequences corresponding to the original signal sequence are determined.

[0068] By determining a plurality of scale parameters corresponding to the signal conversion method and a plurality of time parameters corresponding to the corresponding scale parameters, it is ensured that the original signal sequence can be comprehensively analyzed from the perspectives of multi-scale and time windows, which helps to capture different amplitude characteristics (such as frequency and time-scale characteristics) in the power signal, and helps to reveal key characteristics such as periodicity, transience, and long-term trends in the power signal. By comparing the original signal sequence with these signal comparison sequences, the power signal can be observed from multiple perspectives, avoiding information omission that may be caused by single-scale analysis, improving the comprehensiveness and accuracy of signal feature analysis, and thus helping to provide comprehensive information for the subsequent determination of the type of power signal.

[0069] As an alternative embodiment, determining the amplitude feature sequence corresponding to the original signal sequence includes: determining the amplitude range parameter corresponding to the original signal sequence; determining the amplitude filtering sequence corresponding to the original signal sequence according to the amplitude range parameter corresponding to the original signal sequence; and determining the amplitude feature sequence corresponding to the original signal sequence according to the amplitude filtering sequence and the original signal sequence.

[0070] In this embodiment, the specific steps for determining the amplitude feature sequence corresponding to the original signal sequence are described.

[0071] Among them, the amplitude range parameter is involved. The amplitude range parameter is pre-determined according to the processing target or requirement of the power signal, and is a parameter used to define the effective range of the signal amplitude (such as signal frequency) of the power signal under normal circumstances. For example, the frequency components of partial discharge signals generally gather in the range of 150 kHz to 5 MHz.

[0072] Among them, the amplitude filtering sequence is involved. The amplitude filtering sequence is determined according to the amplitude range parameter and is a filtering sequence used to perform amplitude screening on the original signal sequence.

[0073] In the steps involved in this embodiment, first, the amplitude range parameter corresponding to the original signal sequence is determined. Then, according to the amplitude range parameter corresponding to the original signal sequence, the amplitude filtering sequence corresponding to the original signal sequence is determined. Finally, according to the amplitude filtering sequence and the original signal sequence, the amplitude feature sequence corresponding to the original signal sequence is determined.

[0074] By determining the amplitude range parameter, the key amplitude region of the power signal is clarified, thus enhancing the pertinence and efficiency of power signal processing. By determining the amplitude feature sequence corresponding to the original signal sequence based on the amplitude filtering sequence and the original signal sequence, the power signal components in the original signal sequence that exceed the preset amplitude range are removed, leaving the power signals within the amplitude range, thereby providing a more focused data basis for subsequent signal analysis and feature extraction, and further contributing to improving the accuracy of power signal type determination.

[0075] As an alternative embodiment, determining the first feature parameter corresponding to the first amplitude feature value based on the first amplitude feature value corresponding to the first time point includes: determining the initial time point corresponding to the amplitude feature sequence and the initial feature parameter corresponding to the initial time point; determining the first screening index corresponding to the first amplitude feature value based on the initial feature parameter and the first amplitude feature value corresponding to the first time point; and determining the first feature parameter corresponding to the first amplitude feature value based on the first screening index.

[0076] In this embodiment, the specific steps of determining the first feature parameter corresponding to the first amplitude feature value based on the first amplitude feature value corresponding to the first time point are described.

[0077] Among them, the initial time point is involved. The initial time point is a preset initial time point of the amplitude feature sequence, which can be set to 0 and is the time point before the first time point in the amplitude feature sequence.

[0078] Among them, the initial feature parameter is involved. The initial feature parameter is the feature parameter corresponding to the initial time point in the amplitude feature sequence, which can be set to 0 and is a feature parameter for initializing the amplitude feature sequence. For example, when using a long short-term memory network (LSTM) to process the power signal sequence, a starting point and an initial state are set.

[0079] Among them, the first screening index is involved. The first screening index is an index used to determine the importance of the first amplitude feature value or the initial feature parameter in the amplitude feature sequence for analyzing the power signal features corresponding to the amplitude feature sequence.

[0080] In the steps involved in this embodiment, first, the initial time point corresponding to the amplitude feature sequence and the initial feature parameter corresponding to the initial time point are determined. Then, based on the initial feature parameter and the first amplitude feature value corresponding to the first time point, the first screening index corresponding to the first amplitude feature value is determined. Finally, based on the first screening index, the first feature parameter corresponding to the first amplitude feature value is determined.

[0081] Determining the initial time point corresponding to the amplitude feature sequence and the initial feature parameters corresponding to the initial time point provides a memory starting point for processing the amplitude feature sequence, ensuring that the model (such as the LSTM model) can start learning the features of the amplitude feature sequence from a neutral initial state, avoiding possible biases during the processing, and ensuring the accuracy of signal processing. By determining the first screening index corresponding to the first amplitude feature value based on the initial feature parameters and the first amplitude feature value corresponding to the first time point, it helps the model (such as the LSTM model) to automatically decide which information needs to be stored (such as the forget gate), which needs to be updated (such as the input gate), and which needs to be passed as output to the next time step (such as the output gate) according to the real-time features of the power signal. This screening mechanism helps the model to maintain its long-term memory ability while effectively filtering out noise and irrelevant information when processing complex signal sequences, improving the efficiency and accuracy of signal processing.

[0082] As an alternative embodiment, determining the first feature parameter corresponding to the first amplitude feature value according to the first screening index includes: when the first screening index includes a first retention index, a first update index, and a first weight index, determining the sequence feature parameter corresponding to the first amplitude feature value according to the first retention index and the first update index, where the first retention index is used to represent the retention degree of the initial feature parameters, the first update index is used to represent the update degree of the first amplitude feature value to the initial feature parameters, and the first weight index is used to represent the importance degree of the first amplitude feature value for characterizing the amplitude feature sequence; determining the first feature parameter corresponding to the first amplitude feature value according to the first weight index and the sequence feature parameter corresponding to the first amplitude feature value.

[0083] In this embodiment, the specific steps of determining the first feature parameter corresponding to the first amplitude feature value according to the first screening index are described.

[0084] Among them, the first retention index is involved, and the first retention index is used to represent the retention degree of the initial feature parameters at the current analysis time point. That is, during the signal analysis iteration process, what proportion of the initial feature parameters should be retained. For example, if the first retention index is set to 0.6, it means that the feature parameters at the current time point will retain 60% of the information of the initial feature parameters, which reflects the degree of dependence on the initial state.

[0085] Among them, a first update index is involved. The first update index is used to represent the update amount or influence degree of the first amplitude eigenvalue on the initial feature parameters before update. In iterative signal processing, this can be understood as the amplitude by which the feature parameters at the current time point should be adjusted according to the first amplitude eigenvalue. For example, when the first update index is 0.4, it means that according to the first amplitude eigenvalue, the feature parameters at the current time point will be updated by 40%, reflecting the influence of the new signal features on the current analysis state.

[0086] Among them, a first weight index is involved. The first weight index is used to quantify the relative importance of the first amplitude eigenvalue in constructing the final feature parameters. In a signal processing algorithm, this can be a coefficient used to adjust the contribution size of different signal eigenvalues to the final result. For example, if the partial discharge information carried by the first amplitude eigenvalue is considered more critical than the eigenvalues at other time points, the first weight index can be set to a higher value, such as 0.8, to highlight the importance of this eigenvalue.

[0087] Among them, a sequence feature parameter is involved. The sequence feature parameter is used to represent the parameter that determines the sequence features of the amplitude feature sequence at the first time point and the previous time points according to the first amplitude eigenvalue at the current time point (i.e., the first time point) and in combination with the initial feature parameters.

[0088] In the steps involved in this embodiment, when the first screening index includes a first retention index, a first update index, and a first weight index, first, according to the first retention index and the first update index, the sequence feature parameter corresponding to the first amplitude eigenvalue is determined. Then, according to the first weight index and the sequence feature parameter corresponding to the first amplitude eigenvalue, the first feature parameter corresponding to the first amplitude eigenvalue is determined.

[0089] By determining the sequence feature parameter corresponding to the first amplitude eigenvalue according to the first retention index and the first update index, both historical information can be inherited and a response to new inputs can be made, thus achieving the ability to more accurately capture the dynamic characteristics of power signals while maintaining long-term dependencies. Through the adjustment of the first weight index, the distribution of key features in the amplitude feature sequence can be grasped more precisely, ensuring that key features are fully highlighted and the contributions of irrelevant or interfering features are appropriately weakened when constructing the final feature parameters. Through the comprehensive application of the first retention index, the first update index, and the first weight index, the sequence feature parameter can be dynamically adjusted to ensure that it can accurately reflect the current signal state and will not deviate from the normal trend due to outliers at a single time point, thus ensuring that the most relevant features can be extracted from complex signals, which helps to improve the accuracy of power signal classification.

[0090] As an alternative embodiment, determining the target signal type corresponding to the original signal sequence according to the target characteristic parameters includes: determining the linear transformation parameter and the threshold parameter; determining the linear characteristic parameter corresponding to the original signal sequence according to the linear transformation parameter and the target characteristic parameter; and determining the target signal type corresponding to the original signal sequence according to the linear characteristic parameter and the threshold parameter.

[0091] In this embodiment, the specific steps of determining the target signal type corresponding to the original signal sequence according to the target characteristic parameters are described.

[0092] Among them, the linear transformation parameter is involved. The linear transformation parameter is a parameter used to perform a linear transformation on signal characteristic parameters (such as target characteristic parameters). The linear transformation parameter can be a trained linear transformation matrix. For example, in a fully connected layer, the linear transformation matrix can distinguish which features are crucial for determining the type of power signal and which are irrelevant or negligible.

[0093] Among them, the threshold parameter is involved. The threshold parameter is a constant value added to the output of each power signal classification. The threshold parameter adjusts the baseline of the classification output, enabling the model to better fit the data.

[0094] Among them, the linear characteristic parameter is involved. The linear characteristic parameter is the characteristic parameter obtained by performing a linear transformation on the target characteristic parameter through the linear transformation parameter.

[0095] In the steps involved in this embodiment, first, the linear transformation parameter and the threshold parameter are determined. Then, according to the linear transformation parameter and the target characteristic parameter, the linear characteristic parameter corresponding to the original signal sequence is determined. Finally, according to the linear characteristic parameter and the threshold parameter, the target signal type corresponding to the original signal sequence is determined.

[0096] By determining the linear characteristic parameter corresponding to the original signal sequence according to the linear transformation parameter and the target characteristic parameter, those features that are most important for signal type classification can be identified and weighted, while reducing or ignoring those unimportant or irrelevant features. This enables the model to perform signal type classification based on the optimized feature set, avoiding unnecessary information interference and improving the accuracy and efficiency of classification. The linear characteristic parameter already contains the optimized signal feature information, while the threshold parameter defines a classification boundary, thus helping the model to quickly and accurately determine the specific signal type.

[0097] Based on the above embodiment and alternative embodiment, an alternative implementation manner is provided, which is specifically described below.

[0098] In related technologies, when determining the type of the collected power signal, due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in the technical problem that the determination of the power signal type is inaccurate.

[0099] For the above problems, no effective solution has been proposed yet.

[0100] In view of this, in an alternative embodiment of the present invention, a method for determining the type of a power signal is provided, which can effectively solve the technical problem that due to the diversification and complexity of electromagnetic interference sources, the power signal is interfered by various noises, resulting in inaccurate determination of the power signal type. The following is a specific description.

[0101] S1. Obtain an original signal sequence;

[0102] S2. Determine an amplitude feature sequence corresponding to the original signal sequence, where the amplitude feature sequence includes amplitude feature values corresponding to multiple time points arranged in chronological order;

[0103] Specifically, S2 further includes:

[0104] S21. Determine a signal conversion method corresponding to the original signal sequence;

[0105] S22. Determine multiple scale parameters corresponding to the signal conversion method, and multiple time parameters corresponding to the corresponding scale parameters;

[0106] S23. According to the signal conversion method, determine initial comparison sequences corresponding to the multiple scale parameters respectively;

[0107] S24. According to the initial comparison sequences corresponding to the multiple scale parameters respectively, and the multiple time parameters corresponding to the corresponding scale parameters, determine multiple signal comparison sequences corresponding to the original signal sequence;

[0108] S25. According to the multiple signal comparison sequences and the original signal sequence, determine the amplitude feature sequence corresponding to the original signal sequence.

[0109] For example, wavelet filtering is adopted. Wavelet transform can simultaneously reflect the time-domain and frequency-domain characteristics of a signal, overcome the shortcoming that Fourier transform cannot reflect the local information of a signal, and can effectively remove periodic narrowband interference and background noise interference in partial discharge signals.

[0110] The selection of wavelet basis functions has a significant impact on the results of signal processing. Since each wavelet function has its unique characteristics, different processing results may be obtained when using different wavelet functions for wavelet denoising of the same signal. Similarly, even the same wavelet function will produce different effects when applied to different signals. For example, the signal processed by a wavelet basis function with better symmetry is not prone to phase distortion, while the waveform of the signal processed by a wavelet basis function with better regularity is smoother.

[0111] In summary, when selecting wavelet basis functions, their symmetry and regularity should be considered as much as possible to obtain more effective denoising results. Commonly used wavelet basis functions mainly include Haar wavelet, Daubechies (i.e., db4) wavelet series, Symlets wavelet series, Coiflet wavelet series, and Morlet wavelet, etc. Haar wavelet is one of the earliest used orthogonal wavelet functions and is famous for its compact support. The Daubechies wavelet series is favored for its orthogonality and effective analysis ability. The Symlets wavelet series is an approximately symmetric version based on the Daubechies wavelet. The Coiflet wavelet series is constructed by Daubechies and has better symmetry. The Morlet wavelet has excellent localization ability in the time-frequency domain. Although its orthogonality is weak, it is one of the most commonly used complex-valued wavelet functions.

[0112] Among them, the db4 wavelet basis function can be used for wavelet transform of discrete data. The db4 wavelet basis image represents the detailed part of the signal at different scales and can be used to analyze the local changes and high-frequency components of the signal. The db4 wavelet basis function has a certain smoothness compared with other functions, and can effectively remove the high-frequency noise in the signal while retaining the main features of the signal. Moreover, the db4 wavelet basis function has the characteristics of multi-resolution analysis, can decompose and reconstruct the signal at different scales, which is conducive to capturing the local and global features of the signal. Compared with some complex wavelet basis functions, the db4 wavelet basis function is also computationally efficient and suitable for signal processing in practical applications.

[0113] Perform time-frequency decomposition on the collected original signal (the same as the above original signal sequence) using wavelet transform. Let the original signal be x(t), select a suitable wavelet function ψ(t), and perform continuous wavelet transform (CWT) on x(t):

[0114]

[0115] Where: x(t) is the original signal (same as the above original signal sequence); a is the scale factor (same as the above scale parameter); b is the translation factor (same as the above time parameter); ψ(t) is the wavelet function (same as the above signal conversion method); WT x (a, b) are the wavelet coefficients; t represents the ordinal number;

[0116] By changing the values of a and b, the wavelet coefficients WT x (a, b) can be obtained at different scales and translation positions, and these coefficients constitute the representation of the signal in the time-frequency domain.

[0117] Construct a feature vector based on the wavelet coefficients (same as the above amplitude feature sequence). Calculate statistical features such as the energy, variance, and kurtosis of the wavelet coefficients at different scales, and combine these features into a feature vector. For the energy E i of the wavelet coefficients at scale a ai The calculation formula is:

[0118]

[0119] where i represents the ordinal number;

[0120] Input the constructed feature vector into the trained deep learning model for interference recognition and removal. The deep learning model adopted is an architecture combining a long short-term memory network (LSTM) and a fully connected layer. The LSTM layer can effectively handle the long-term dependencies in time series data and has good adaptability to the time-varying characteristics of partial discharge signals and interference signals.

[0121] Specifically, S2 also includes:

[0122] S201, determine the amplitude range parameter corresponding to the original signal sequence;

[0123] S202, determine the amplitude filtering sequence corresponding to the original signal sequence according to the amplitude range parameter corresponding to the original signal sequence;

[0124] S203, determine the amplitude feature sequence corresponding to the original signal sequence according to the amplitude filtering sequence and the original signal sequence.

[0125] For example, band-pass filtering is adopted. The frequency components of partial discharge signals generally concentrate in the range of 150 kHz to 5 MHz (same as the above amplitude range parameter), while the frequency band of background noise is widely distributed. A band-pass filter can be used to filter out the noise components with frequencies outside the partial discharge frequency. According to the difference in impulse response, band-pass filters can be divided into two types: Finite Impulse Response (FIR) filters and Infinite Impulse Response (IIR) filters. Since FIR can obtain strict linear phase stability, by using FIR filter band-pass filtering, the noise signals outside the passband above 5 MHz (same as the above amplitude range parameter) can be effectively suppressed, which is mainly applicable to random white noise.

[0126] S3. In the order of time points, based on the first amplitude eigenvalue corresponding to the first time point, determine the first characteristic parameter corresponding to the first amplitude eigenvalue, and based on the first characteristic parameter and the second amplitude eigenvalue corresponding to the second time point, determine the second characteristic parameter corresponding to the second amplitude eigenvalue, until all the amplitude eigenvalues in the amplitude eigenvalue sequence are processed to obtain the target characteristic parameter.

[0127] Specifically, S3 further includes:

[0128] S31. Determine the initial time point corresponding to the amplitude eigenvalue sequence and the initial characteristic parameter corresponding to the initial time point.

[0129] S32. Based on the initial characteristic parameter and the first amplitude eigenvalue corresponding to the first time point, determine the first screening index corresponding to the first amplitude eigenvalue.

[0130] S33. Based on the first screening index, determine the first characteristic parameter corresponding to the first amplitude eigenvalue.

[0131] S34. When the first screening index includes the first retention index, the first update index, and the first weight index, based on the first retention index and the first update index, determine the sequence characteristic parameter corresponding to the first amplitude eigenvalue, where the first retention index is used to represent the retention degree of the initial characteristic parameter, the first update index is used to represent the update degree of the first amplitude eigenvalue to the initial characteristic parameter, and the first weight index is used to represent the importance degree of the first amplitude eigenvalue for characterizing the amplitude eigenvalue sequence.

[0132] S35. Based on the first weight index and the sequence characteristic parameter corresponding to the first amplitude eigenvalue, determine the first characteristic parameter corresponding to the first amplitude eigenvalue.

[0133] For example, a deep learning architecture combining a long short-term memory network (LSTM) and a fully connected layer is constructed. The LSTM layer, as the core component of the model, has a key component in its internal structure, including an input gate, a forget gate, an output gate, and a memory cell, etc., which can effectively handle the long-term dependence relationship in time series data.

[0134] Let the input of the LSTM layer be x t (the values of the feature vector at different time steps, that is, the values of the preprocessed signal features at different time steps), the hidden state be h t-1 , and the cell state be c t-1 . The input gate of the LSTM layer controls the degree to which the current input information enters the memory cell. The update formula of the input gate is as follows:

[0135] i t = σ(W i x t + U i h t-1 + b i )

[0136] Among them, i t represents the update exponent (the same as the first update exponent above); σ is the sigmoid function; W i and U i are the weight matrices of the input gate, and b i is the bias vector.

[0137] The forget gate of the LSTM layer is responsible for determining which information in the memory cell of the previous moment needs to be retained or forgotten. The calculation formula of the forget gate is:

[0138] f t = σ(W f x t + U f h t-1 + b f )

[0139] Among them, f t represents the retention exponent (the same as the first retention exponent above), W f and U f are the weight matrices of the forget gate; b f is the bias vector of the forget gate.

[0140] The output gate calculates the proportion of the output information according to the current input, the hidden state of the previous moment, and the weight bias. The calculation formula of the output gate is:

[0141] o t = σ(W o x t + U o h t-1 + bo )

[0142] Where: o t represents the weight index (same as the above first weight index), W o and U o are the weight matrices of the output gate, and b o is the bias vector of the output gate.

[0143] The memory cell update formula is:

[0144] c t = f t ⊙ c t-1 + i t ⊙ cg t

[0145] Where, c t represents the cell state (same as the above sequence feature parameter), cg t represents the candidate hidden state, that is, the candidate output result, and ⊙ represents element-wise multiplication.

[0146] cg t The calculation formula of

[0147] cg t = tanh(W g x t + U g h t-1 + b g )

[0148] Where, W g and U g represent the corresponding weight matrices, and b g represents the corresponding bias vector.

[0149] h t The calculation formula of

[0150] h t = o t ⊙ tanh(c t )

[0151] Where, h t represents the hidden state (same as the above first feature parameter, second feature parameter or target feature parameter)

[0152] Through the above formulas, the effective management and update of long-term dependence information can be achieved.

[0153] S4. According to the target feature parameter, determine the target signal type corresponding to the original signal sequence.

[0154] Specifically, S4 further includes:

[0155] S41, Determine the linear transformation parameters and threshold parameters;

[0156] S42, According to the linear transformation parameters and the target feature parameters, determine the linear feature parameters corresponding to the original signal sequence;

[0157] S43, According to the linear feature parameters and the threshold parameters, determine the target signal type corresponding to the original signal sequence.

[0158] Connect a fully connected layer after the LSTM layer. The role of the fully connected layer is to further integrate and classify the features output by the LSTM layer. Let the input of the fully connected layer be the last hidden state h L , The calculation formula for the output yout of the fully connected layer is:

[0159] yout = W fc h L + b fc

[0160] Among them, yout represents the output of the fully connected layer (the same as the above target signal type); W fc is the weight matrix of the fully connected layer (the same as the above linear transformation parameters); h L represents the last hidden state (the same as the above target feature parameters); b fc is the bias vector (the same as the above threshold parameters).

[0161] By adjusting the weights and biases of the fully connected layer, the model can accurately output the classification results of the signals. To prevent the model from overfitting, a dropout layer is added to the model. During the training process, with a certain probability (such as 0.5), the outputs of some neurons are randomly set to 0, so that the model does not overly rely on certain specific neurons during the training process and enhances the generalization ability of the model.

[0162] During the training phase, input the training set data into the constructed model. Use an appropriate loss function to measure the difference between the model prediction results and the true labels. The calculation formula for the cross-entropy loss function used is:

[0163]

[0164] Among them, L represents the loss value, N is the number of samples; y i is the true label; is the model prediction result.

[0165] Use the Adam optimization algorithm to update the model parameters. It adaptively adjusts the learning rate of each parameter and updates the parameters by calculating the first-order moment estimate and second-order moment estimate of the gradient during the training process. In each iteration, calculate the gradient g of the parametert , then update the first - order moment estimate and the second - order moment estimate of the gradient, and the formulas are as follows:

[0166] m t =β 1 m t-1 +(1 - β 1 )g t

[0167]

[0168] Among them, m t is the first - order moment estimate value, v t is the second - order moment estimate value, β 1 and β 2 are the decay rates.

[0169] Then perform bias correction on the first - order moment and the second - order moment, and the formulas are as follows:

[0170]

[0171] Among them, is the first - order moment correction value; is the second - order moment correction value; and are the correction coefficients.

[0172] Finally, update the parameters, and the formula is as follows:

[0173]

[0174] Among them, θ t+1 represents the updated parameter, θ t represents the parameter before update; α is the learning rate; ∈ is a small constant to prevent division by zero.

[0175] During the training process, set the number of training epochs to 100 and the batch size to 32. After training each batch of data, calculate the performance metrics such as the loss value and accuracy of the model on the validation set. By observing the change trend of the validation set performance metrics, when it is found that the validation set loss value no longer decreases or the accuracy no longer improves, it indicates that the model may have overfitting phenomenon. At this time, the early stopping method can be used to stop the training to avoid the decline of the model performance caused by over - training. At the same time, adjust the hyperparameters of the model according to the validation set performance.

[0176] After the training is completed, use the test set to perform the final evaluation of the model. Calculate the performance metrics such as the accuracy, recall rate, and F1 - value of the model on the test set. The accuracy calculation formula is:

[0177]

[0178] Among them, Accuracy represents the accuracy rate; TP is the true positive, which is the number of correctly predicted partial discharge signals; TN is the true negative, which is the number of correctly predicted interference signals; FP is the false positive, which is the number of interference signals misjudged as partial discharge signals; FN is the false negative, which is the number of partial discharge signals misjudged as interference signals.

[0179] The formula for calculating the recall rate is:

[0180]

[0181] Among them, Recall represents the recall rate.

[0182] The formula for calculating Precision is:

[0183]

[0184] Among them, Precision represents the precision rate.

[0185] The formula for calculating the F1 value is:

[0186]

[0187] Among them, F1 represents the F1 value, which is the harmonic mean of the precision rate and the recall rate;

[0188] The performance of the model is comprehensively evaluated through performance indicators in actual applications. After the model performance meets the expected requirements, it is applied to the actual cable partial discharge detection system. During the application process, the cable operation signals are collected in real time. After data preprocessing and wavelet threshold denoising, they are input into the trained model. The model outputs the classification results and confidence scores of the signals. A reasonable threshold is set according to the confidence score. When the confidence score is higher than the threshold, it is considered that the prediction result of the model has high reliability. If the confidence score is greater than 0.8, the output result of the model is used as the final judgment basis. If the output is a partial discharge signal, an alarm is issued in a timely manner and further analysis is carried out; if the output is an interference signal, the signal can be ignored, thereby effectively improving the accuracy and reliability of cable partial discharge detection.

[0189] Through the above steps, a specific example is as follows:

[0190] For example, the acquired original signal data (the same as the above original signal sequence) is preprocessed. The specific steps include:

[0191] A1. Conduct data collection and preprocessing:

[0192] Collect signals during the operation of the cable using high-precision sensors to ensure that the collected signals have sufficient sampling frequency and resolution to capture partial discharge signals and possible high-frequency pulse interference signals. Normalize the collected original signals to make their amplitude ranges within appropriate intervals, facilitating subsequent wavelet transform and model training.

[0193] A2. Wavelet threshold denoising:

[0194] Denoise the signal according to the different properties of the wavelet coefficients obtained by decomposing the signal and noise in the wavelet domain, and use an appropriate threshold to compress or non-linearly process the wavelet signal containing noise.

[0195] Assume that the model of the noisy signal is as follows:

[0196] f(n) = s(n) + δ·e(n)

[0197] Where, f(n) is the noisy signal; s(n) is the true signal; δ is the standard deviation of the noise coefficient; e(n) is the noise signal.

[0198] A3. Perform wavelet decomposition on the noisy signal:

[0199] Select appropriate wavelet basis functions and the number of wavelet decomposition layers according to the characteristics of the signal data, perform wavelet decomposition on the signal, and the obtained wavelet coefficients are:

[0200] ω s (j,k) = ω x (j,k) + ω n (j,k)

[0201] Where, ω s (j,k) represents the wavelet coefficient of the noisy signal on the j-th layer; ω x (j,k) represents the wavelet coefficient of the pure signal on the j-th layer; ω n (j,k) represents the wavelet coefficient of the noise signal on the j-th layer.

[0202] A4. Threshold denoising processing:

[0203] Select appropriate wavelet thresholds and threshold functions to process the wavelet coefficients after wavelet decomposition, so that the coefficients of the noise in the signal are set to zero, extract the coefficients of the useful part in the signal, obtain the new wavelet coefficients, and achieve the purpose of denoising. The denoising results of two groups of overhead lines in different states are subjected to SureShrink threshold wavelet denoising using the soft threshold (soft) function. This denoising technique not only successfully retains the basic shape and dynamic trend of the signal, but also effectively suppresses the noise components. This denoising method also largely maintains the characteristics of partial discharge while denoising.

[0204] A5. Deep learning model training and application:

[0205] A deep learning model is constructed based on LSTM and fully connected layers, and the signal obtained after wavelet threshold denoising is used as the input of the deep learning model. The LSTM layer can automatically learn the time series features of the signal, capture the variation laws and dependence relationships of partial discharge signals and interference signals in the time dimension, and the connection layer further integrates and classifies the features output by the LSTM layer.

[0206] In the initial stage of deep learning model training and application, the quality and preparation of data play a key role. Abundant and diverse data samples are collected from a large number of actual cable monitoring scenarios, and these samples cover partial discharge signals and various interference signals under different cable types, operating environments, and load conditions. The collected raw data is strictly screened and cleaned to remove obvious abnormal or incorrect data points to ensure the accuracy and reliability of the data. Before the data is input into the model, the data is normalized. For features such as the amplitude and frequency of partial discharge signals and interference signals, standardization or normalization methods are used to map them to a specific interval range, making different features comparable, avoiding adverse effects on model training caused by data dimension differences, and providing a good data basis for model training. The formula is as follows:

[0207]

[0208] Among them, X norm represents the normalized value of feature X; X min is the minimum value of feature X; X max is the maximum value of feature X.

[0209] Through the above optional implementation manners, at least the following beneficial effects can be achieved:

[0210] (1) Compared with the related technologies, the present invention extracts time-related amplitude feature information from the original signal sequence, which helps to more intuitively reflect the change of signal intensity. By sequentially analyzing the features of multiple amplitude feature values in the amplitude feature sequence according to the time point order, the changes of power signals at each time point can be effectively captured and analyzed, and the overall features of the amplitude feature sequence can be accurately identified, so that the features of power signals can be accurately captured. Using the target feature parameters of the extracted power signals for classification helps to accurately identify the specific types of power signals, thus ensuring the accuracy and reliability of power signal type determination in complex environments such as the presence of high-frequency pulse interference. Furthermore, it solves the technical problem that due to the diversification and complexity of electromagnetic interference sources, power signals are interfered by various noises, resulting in inaccurate determination of power signal types.

[0211] (2) Compared with the related technologies, the present invention provides diverse perspectives of signals in different domains or scales by determining multiple signal comparison sequences, which can help identify subtle differences and interference patterns in power signals, providing a multi-dimensional analysis basis for the determination of amplitude feature sequences. By comparing the original signal sequence with the signal comparison sequences, the effective information in the power signal can be highlighted, while the interference components can be suppressed or removed, thus helping to identify and quantify the amplitude features of the power signal, contributing to subsequent power signal classification. It can highlight the effective information in the signal while suppressing or removing the interference components.

[0212] (3) Compared with the related technologies, the present invention ensures that the original signal sequence can be comprehensively analyzed from the perspectives of multi-scale and time window by determining multiple scale parameters corresponding to the signal conversion method and multiple time parameters corresponding to the corresponding scale parameters, which helps capture different amplitude features in the power signal and reveals key features such as periodicity, transience, and long-term trends in the power signal. By comparing the original signal sequence with these signal comparison sequences, the power signal can be observed from multiple angles, avoiding information omission caused by single-scale analysis, improving the comprehensiveness and accuracy of signal feature analysis, and thus contributing to providing comprehensive information for the subsequent determination of the type of power signal.

[0213] (4) Compared with the related technologies, the present invention determines the sequence feature parameters corresponding to the first amplitude feature value according to the first retention index and the first update index, which can not only inherit historical information but also respond to new inputs, thus achieving the ability to more accurately capture the dynamic characteristics of power signals while maintaining long-term dependencies. By adjusting the first weight index, the distribution of key features in the amplitude feature sequence can be more precisely grasped, ensuring that key features are fully highlighted while the contributions of irrelevant or interfering features are appropriately weakened when constructing the final feature parameters. Through the comprehensive application of the first retention index, the first update index, and the first weight index, the sequence feature parameters can be dynamically adjusted to ensure that it can accurately reflect the current signal state and will not deviate from the normal trend due to outliers at a single time point, thus ensuring that the most relevant features can be extracted from complex signals, contributing to improving the accuracy of power signal classification.

[0214] (5) Compared with the related technologies, the present invention combines time-frequency analysis and deep learning. The signal after band-pass filtering and wavelet filtering is input into the LSTM and fully connected layer models after wavelet coefficient feature extraction. This process synthesizes the processing results of different filtering technologies and multi-scale wavelet coefficient features, forming a unique multi-stage fusion processing system. Compared with traditional single filtering or simple threshold discrimination, it can capture signal features more comprehensively, greatly improve the interference recognition accuracy, and effectively cope with interference in complex electromagnetic environments.

[0215] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0216] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of the various embodiments of the present invention.

[0217] Embodiment 2

[0218] According to an embodiment of the present invention, there is also provided a device for implementing the above-mentioned power signal type determination method. Figure 2 It is a structural block diagram of the power signal type determination device according to an embodiment of the present invention, as Figure 2 shown. The device includes: an acquisition module 202, a first determination module 204, a second determination module 206, and a third determination module 208. The device will be described in detail below.

[0219] The acquisition module 202 is used to acquire an original signal sequence; the first determination module 204 is connected to the above-mentioned acquisition module 202 and is used to determine an amplitude feature sequence corresponding to the original signal sequence, where the amplitude feature sequence includes a plurality of amplitude feature values arranged in the order of time points; the second determination module 206 is connected to the above-mentioned first determination module 204 and is used to determine, in the order of time points, a first feature parameter corresponding to the first amplitude feature value corresponding to the first time point, and determine a second feature parameter corresponding to the second amplitude feature value according to the first feature parameter and the second amplitude feature value corresponding to the second time point, until all the amplitude feature values in the amplitude feature sequence are processed to obtain a target feature parameter; the third determination module 208 is connected to the above-mentioned second determination module 206 and is used to determine a target signal type corresponding to the original signal sequence according to the target feature parameter.

[0220] It should be noted here that the above-mentioned acquisition module 202, the first determination module 204, the second determination module 206, and the third determination module 208 correspond to steps S102 to S108 in the method for determining the type of power signal. The instances and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the content disclosed in the above-mentioned Embodiment 1.

[0221] Embodiment 3

[0222] According to another aspect of the embodiments of the present invention, an electronic device is further provided, including: a processor; a memory for storing instructions executable by the processor, wherein the processor is configured to execute the instructions to implement the method for determining the type of power signal according to any one of the above.

[0223] Embodiment 4

[0224] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is further provided. When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the method for determining the type of power signal according to any one of the above.

[0225] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0226] In the above embodiments of the present invention, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0227] In the several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of the units or modules can be in electrical or other forms.

[0228] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0229] In addition, in each embodiment of the present invention, each functional unit may be integrated into a processing unit, may exist physically alone for each unit, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0230] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disc that can store program codes.

[0231] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for determining the type of a power signal, characterized in that: include: Obtaining the original signal sequence; Determine an amplitude characteristic sequence corresponding to the original signal sequence, wherein the amplitude characteristic sequence includes amplitude characteristic values ​​corresponding to a plurality of time points arranged in chronological order; According to the time point sequence, based on the first amplitude characteristic value corresponding to the first time point, a first characteristic parameter corresponding to the first amplitude characteristic value is determined, and based on the first characteristic parameter and the second amplitude characteristic value corresponding to the second time point, a second characteristic parameter corresponding to the second amplitude characteristic value is determined, until the multiple amplitude characteristic values ​​in the amplitude characteristic sequence are processed, and the target characteristic parameter is obtained; A target signal type corresponding to the original signal sequence is determined according to the target characteristic parameter.

2. The method according to claim 1, characterized in that The determining of the amplitude characteristic sequence corresponding to the original signal sequence comprises: Determining a signal conversion mode corresponding to the original signal sequence; According to the signal conversion method, a plurality of signal comparison sequences corresponding to the original signal sequence are determined, wherein the plurality of signal comparison sequence lists are sequences used for comparison with the original signal sequence; Based on the multiple signal comparison sequences and the original signal sequence, an amplitude feature sequence corresponding to the original signal sequence is determined.

3. The method according to claim 2, characterized in that The step of determining a plurality of signal comparison sequences corresponding to the original signal sequence according to the signal conversion method includes: Determining a plurality of scale parameters corresponding to the signal conversion mode and a plurality of time parameters corresponding to the corresponding scale parameters; Determining initial comparison sequences corresponding to the plurality of scale parameters respectively according to the signal conversion method; According to the initial comparison sequences corresponding to the multiple scale parameters respectively and the multiple time parameters corresponding to the corresponding scale parameters, multiple signal comparison sequences corresponding to the original signal sequence are determined.

4. The method according to claim 1, characterized in that: The determining of the amplitude characteristic sequence corresponding to the original signal sequence comprises: Determining an amplitude range parameter corresponding to the original signal sequence; Determining an amplitude filtering sequence corresponding to the original signal sequence according to an amplitude range parameter corresponding to the original signal sequence; According to the amplitude filtering sequence and the original signal sequence, an amplitude feature sequence corresponding to the original signal sequence is determined.

5. The method according to claim 1, characterized in that The determining, based on the first amplitude characteristic value corresponding to the first time point, a first characteristic parameter corresponding to the first amplitude characteristic value comprises: Determining an initial time point corresponding to the amplitude characteristic sequence, and an initial characteristic parameter corresponding to the initial time point; Determining a first screening index corresponding to the first amplitude characteristic value according to the initial characteristic parameter and the first amplitude characteristic value corresponding to the first time point; A first characteristic parameter corresponding to the first amplitude characteristic value is determined according to the first screening index.

6. The method according to claim 5, characterized in that The determining, according to the first screening index, a first characteristic parameter corresponding to the first amplitude characteristic value comprises: In the case where the first screening index includes a first retention index, a first update index and a first weight index, a sequence characteristic parameter corresponding to the first amplitude characteristic value is determined according to the first retention index and the first update index, wherein the first retention index is used to indicate the degree of retention of the initial characteristic parameter, the first update index is used to indicate the degree of update of the initial characteristic parameter by the first amplitude characteristic value, and the first weight index is used to indicate the importance of the first amplitude characteristic value for characterizing the amplitude characteristic sequence; A first characteristic parameter corresponding to the first amplitude characteristic value is determined according to the first weight index and the sequence characteristic parameter corresponding to the first amplitude characteristic value.

7. The method according to any one of claims 1 to 6, characterized in that: The step of determining the target signal type corresponding to the original signal sequence according to the target characteristic parameter includes: Determine linear transformation parameters and threshold parameters; Determining a linear characteristic parameter corresponding to the original signal sequence according to the linear transformation parameter and the target characteristic parameter; The target signal type corresponding to the original signal sequence is determined according to the linear characteristic parameter and the threshold parameter.

8. A device for determining the type of electric power signal, characterized in that: include: An acquisition module, used for acquiring an original signal sequence; A first determination module is used to determine an amplitude characteristic sequence corresponding to the original signal sequence, wherein the amplitude characteristic sequence includes a plurality of amplitude characteristic values ​​arranged in order of time points; A second determination module is used to determine, in a time point sequence, a first characteristic parameter corresponding to the first amplitude characteristic value according to the first amplitude characteristic value corresponding to the first time point, and determine a second characteristic parameter corresponding to the second amplitude characteristic value according to the first characteristic parameter and the second amplitude characteristic value corresponding to the second time point, until the multiple amplitude characteristic values ​​in the amplitude characteristic sequence are processed and the target characteristic parameter is obtained; The third determination module is used to determine the target signal type corresponding to the original signal sequence according to the target characteristic parameter.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method for determining the type of electric power signal according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method for determining the type of a power signal as claimed in any one of claims 1 to 7.

Images