Line fault detection method based on polynomial model

Abstract

The application relates to the technical field of electric power, and discloses a line fault detection method based on a polynomial model, which comprises the following steps: determining a plurality of time-varying coefficients by taking data points at each moment in a historical distortion signal and data points at each moment in an experimental restoration signal as the basis to construct a polynomial model; and substituting a data point at the tth moment in a current distortion signal into the polynomial model to obtain a data point at the tth moment in a target restoration signal, so that the time domain of the current distortion signal is restored, the method can effectively eliminate the nonlinear influence of a mutual inductor on the time domain of a measured signal, accurately restore real fault information, and thus improve the detection precision of hidden faults of an overhead line.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, and particularly relates to a line fault detection method based on a polynomial model. BACKGROUND

[0002] With the continuous development of the power system, as a key component of the power supply network, the safety and reliability of the distribution overhead line are increasingly valued. However, in the actual operation of the distribution overhead line, it is influenced by external environment and line itself, resulting in frequent faults, such as high resistance grounding fault and hidden fault. The traditional fault detection method mainly relies on direct measurement and analysis of the line electrical quantity, but in the complex electromagnetic environment, the signal often appears nonlinear distortion, which makes it extremely difficult to accurately extract the fault characteristics, and further leads to low detection accuracy of the hidden fault.

[0003] As a commonly used signal acquisition device in overhead lines, the non-rational characteristics of the mutual inductor have nonlinear influence on the time domain of the measured signal, which will cause distortion of the fault signal, so that the detection system cannot accurately restore the real fault information, thereby affecting the accuracy of fault detection. SUMMARY

[0004] Therefore, it is necessary to propose a line fault detection method based on a polynomial model to effectively eliminate the nonlinear influence of the mutual inductor on the time domain of the measured signal, accurately restore the real fault information, and improve the detection accuracy of the hidden fault of the overhead line.

[0005] To achieve the above purpose, in a first aspect, the present application provides a line fault detection method based on a polynomial model, which comprises:

[0006] obtaining the current distorted signal and the historical distorted signal of the mutual inductor, and the experimental restoration signal corresponding to the historical distorted signal;

[0007] determining a plurality of time-varying coefficients according to the data points at each time in the historical distorted signal and the data points at each time in the experimental restoration signal, to construct a polynomial model;

[0008] substituting the data point at the tth time in the current distorted signal into the polynomial model to obtain the data point at the tth time in the target restoration signal; wherein t takes an integer greater than 0 in turn until n is equal to the total number of times of the current distorted signal, to obtain the target restoration signal;

[0009] performing multi-feature extraction on the target restoration signal to obtain a signal feature vector, and inputting the signal feature vector into a preset machine learning model to obtain a fault classification detection result.

[0010] Optionally, substituting the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal includes:

[0011] Using the expression of the polynomial model The target reconstruction signal is obtained;

[0012] in, Let be the data point at time t in the target restored signal. For the (n+1)th time-varying coefficient, The total number of time-varying coefficients. Let be the data point at time t in the current distorted signal.

[0013] Optionally, determining multiple time-varying coefficients based on data points at various times in the historical distorted signal and data points at various times in the experimentally reconstructed signal to construct the polynomial model includes:

[0014] The historical distortion signal and the experimental restored signal are standardized respectively to obtain the standard distortion signal and the standard restored signal;

[0015] Using the least squares method, multiple time-varying coefficients are determined based on the data points at each time step in the standard distorted signal and the data points at each time step in the standard restored signal.

[0016] The polynomial model is constructed based on each time-varying coefficient.

[0017] Optionally, the standardization process performed on the historical distortion signal and the experimental restored signal to obtain a standard distortion signal and a standard restored signal includes:

[0018] Determine the first mean and first standard deviation of the historical distorted signal, and determine the second mean and second standard deviation of the experimentally restored signal;

[0019] Based on the first mean and the first standard deviation, the data points at each time point in the historical distortion signal are standardized to obtain the standard distortion signal. Based on the second mean and the second standard deviation, the data points at each time point in the experimental restored signal are standardized to obtain the standard restored signal.

[0020] Optionally, after determining multiple time-varying coefficients based on data points at various times in the historical distorted signal and data points at various times in the experimentally reconstructed signal to construct a polynomial model, the method further includes:

[0021] Fourier transforms are performed on the current distorted signal, the historical distorted signal, and the experimental restored signal, respectively, to obtain the current frequency domain distorted signal, the historical frequency domain distorted signal, and the experimental frequency domain restored signal;

[0022] Based on the historical frequency domain distortion signal and the experimental frequency domain restored signal, a frequency response transfer function is constructed.

[0023] The frequency response transfer function is transformed to obtain a frequency domain form function, and the target frequency domain restored signal is determined based on the amplitude attenuation coefficient and phase offset of the frequency domain form function, as well as the amplitude and phase of each frequency in the current frequency domain distorted signal.

[0024] Perform an inverse Fourier transform on the target frequency domain reconstructed signal to obtain the initial reconstructed signal;

[0025] The initial restored signal is used as the current distortion signal.

[0026] Optionally, the expression for the frequency response transfer function is:

[0027] ;

[0028] in, ;

[0029] In the above formula, The frequency response transfer function is... For the Laplace operator, The transfer coefficient of the (m+1)th molecule. The total number of numerator or denominator transfer coefficients. The transitivity coefficient of the (m+1)th denominator. The imaginary unit, Pi For frequency.

[0030] Optionally, the expression for the frequency domain form function is:

[0031] ;

[0032] in, For the frequency domain form function, The amplitude attenuation coefficient at frequency f is... It is a natural constant. The imaginary unit, The phase offset is the frequency f.

[0033] Optionally, determining the target frequency domain reconstructed signal based on the amplitude attenuation coefficient and phase offset of the frequency domain form function, and the amplitude and phase of each frequency in the current frequency domain distorted signal, includes:

[0034] Using formula Determine the initial frequency domain reconstructed signal;

[0035] in, The frequency domain component of the k-th frequency in the target frequency domain reconstructed signal. For the kth frequency, Let x be the amplitude of the k-th frequency in the current frequency domain distorted signal. For frequency f k The amplitude attenuation coefficient, It is a natural constant. The imaginary unit, The phase of the k-th frequency in the current frequency domain distorted signal. For frequency f k The phase offset.

[0036] Optionally, a Fourier transform is performed on the current distorted signal to obtain the current frequency domain distorted signal, including:

[0037] Perform a Fourier transform on the current distorted signal to obtain the initial frequency domain distorted signal;

[0038] Based on the reconstructed signal from the experiment, determine the maximum frequency;

[0039] The cutoff frequency is determined based on the maximum frequency and the preset safety factor;

[0040] Based on the cutoff frequency, the initial frequency domain distortion signal is frequency-filtered to obtain the current frequency domain distortion signal.

[0041] Optionally, the step of performing multi-feature extraction on the target restored signal to obtain a signal feature vector includes:

[0042] Perform wavelet transform on the target restored signal to obtain the first time-frequency diagram;

[0043] Perform a short-time Fourier transform on the target restored signal to obtain a second time-frequency diagram;

[0044] Multi-feature extraction is performed on the first time-frequency graph and the second time-frequency graph to obtain the spectrum center frequency, bandwidth, short-time energy and time-domain waveform change.

[0045] The center frequency of the spectrum, the bandwidth, the short-time energy, and the time-domain waveform change are used as the signal feature vector.

[0046] To achieve the above objectives, in a second aspect, the present invention provides a line fault detection device based on a polynomial model, the device comprising:

[0047] The acquisition module is used to acquire the current distortion signal and historical distortion signal of the current transformer, as well as the experimental reconstruction signal corresponding to the historical distortion signal;

[0048] The determination and construction module is used to determine multiple time-varying coefficients based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimental restored signal, so as to construct a polynomial model.

[0049] The module is used to substitute the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal; where t takes the values ​​of integers greater than 0 in sequence until n is equal to the total number of times in the current distorted signal, thus obtaining the target restored signal;

[0050] The extraction and prediction module is used to extract multiple features from the target restored signal to obtain a signal feature vector, and input the signal feature vector into a preset machine learning model to obtain a fault classification detection result.

[0051] To achieve the above objectives, the present invention provides, in a third aspect, a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the line fault detection method based on a polynomial model as described in any one of the first aspects.

[0052] To achieve the above objectives, the present invention provides a computer device in a fourth aspect, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor causes the processor to perform the line fault detection method based on a polynomial model as described in any one of the first aspects.

[0053] The present invention provides the following advantages: The method acquires the current distortion signal and historical distortion signal of the transformer, as well as the experimentally reconstructed signal corresponding to the historical distortion signal. Then, based on the data points at each time point in the historical distortion signal and the experimentally reconstructed signal, multiple time-varying coefficients are determined to construct a polynomial model. The data point at time t in the current distortion signal is then substituted into the polynomial model to obtain the data point at time t in the target reconstructed signal, where t takes values ​​greater than 0, up to n equal to the total number of times in the current distortion signal. Finally, multiple features are extracted from the target reconstructed signal to obtain the signal... The signal feature vector is input into a preset machine learning model to obtain the fault classification and detection result. That is, by determining multiple time-varying coefficients based on the data points at each time moment in the historical distorted signal and the data points at each time moment in the experimentally restored signal, a polynomial model is constructed. Then, the data point at time t in the current distorted signal is substituted into the polynomial model to obtain the data point at time t in the target restored signal, so as to restore the time domain of the current distorted signal. This method can effectively eliminate the nonlinear influence of the current transformer on the time domain of the measured signal, accurately restore the real fault information, and thus improve the detection accuracy of hidden faults in overhead lines. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] in:

[0056] Figure 1 This is a schematic diagram of the line fault detection method based on a polynomial model in the embodiments of this application;

[0057] Figure 2 This is a schematic diagram of a line fault detection device based on a polynomial model in an embodiment of this application.

[0058] Figure 3 This is a diagram showing the internal structure of a computer device in some embodiments. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] With the continuous development of power systems, the safety and reliability of overhead distribution lines, as a key component of the power supply network, are receiving increasing attention. However, in actual operation, overhead distribution lines are affected by both external environmental factors and inherent line characteristics, leading to frequent faults such as high-resistance grounding faults and hidden faults. Traditional fault detection methods mainly rely on the direct measurement and analysis of line electrical quantities. However, in complex electromagnetic environments, signals often exhibit nonlinear distortion, making accurate extraction of fault characteristics extremely difficult, resulting in low accuracy in detecting hidden faults.

[0061] As a commonly used signal acquisition device in overhead lines, the irrational characteristics of instrument transformers have a nonlinear effect on the time domain of the measured signal, which can lead to distortion of fault signals. This makes it impossible for the detection system to accurately reproduce the true fault information, thus affecting the accuracy of fault detection.

[0062] To address the aforementioned issues, this application proposes a line fault detection method based on a polynomial model, which can effectively eliminate the nonlinear influence of the instrument transformer on the time domain of the measured signal, accurately restore the true fault information, and thus improve the detection accuracy of hidden faults in overhead lines. The specific implementation principle will be described in detail in the following embodiments.

[0063] It should be noted that this application describes the nonlinear effects of the non-ideal characteristics of the instrument transformer on the measured signal in three stages: the first stage is the nonlinear effects of the non-ideal characteristics of the instrument transformer on the amplitude and phase of the measured signal, which is generally caused by the characteristics of the instrument transformer such as the core permeability and winding inductance; the second stage is the nonlinear effects of the non-ideal characteristics of the instrument transformer on the time domain of the measured signal, which is generally caused by the characteristics of the core saturation and hysteresis; and the third stage is the dynamic nonlinear effects of the non-ideal characteristics of the instrument transformer on the time domain of the measured signal, which is generally caused by the characteristics of the instrument transformer such as temperature drift and core aging.

[0064] This application provides a line fault detection method based on a polynomial model in its first aspect.

[0065] Please see Figure 1 This is a schematic diagram of a line fault detection method based on a polynomial model in an embodiment of this application. The method includes:

[0066] Step 110: Obtain the current distortion signal and historical distortion signal of the current transformer, as well as the experimental restoration signal corresponding to the historical distortion signal.

[0067] Here, the signal can be an electrical signal, including but not limited to current signals, voltage signals, etc.

[0068] It should be noted that the current distortion signal of the current transformer can be obtained by measuring the current and / or voltage in the line in real time; the historical distortion signal can be obtained by inputting the experimental restoration signal into the current transformer and then outputting the signal from the current transformer; the experimental restoration signal can be obtained by the operator using a signal generator.

[0069] Regarding the generation method of the experimental restoration signal, in some embodiments, the operating frequency range of the current transformer can be obtained, the signal generator can be controlled to generate a signal, and within the operating frequency range, the frequency of the generated signal can be adjusted sequentially by a preset step size to obtain a signal with multiple frequencies, which is the experimental restoration signal.

[0070] Regarding the value of the operating frequency range, in some embodiments, this application preferably sets the operating frequency range to 0.1Hz to 10kHz.

[0071] Step 120: Based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimentally restored signal, determine multiple time-varying coefficients to construct a polynomial model.

[0072] Regarding the determination of multiple time-varying coefficients, in some embodiments, a multi-order polynomial can be used. Based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimentally restored signal, multiple coefficients of the multi-order polynomial are determined, and all multiple coefficients are used as time-varying coefficients to obtain multiple time-varying coefficients. The order of the multi-order polynomial can be obtained and preset by the operator based on a large amount of experience, experiments or statistics. Of course, it can also be preset by the operator according to actual needs.

[0073] In some embodiments, multiple time-varying coefficients can be substituted into a multi-order polynomial to obtain the constructed polynomial model.

[0074] It should be noted that the non-ideal characteristics of the current transformer not only have a non-linear effect on the time domain of the measured signal, but also on the amplitude and phase of the measured signal. This non-linear effect on amplitude and phase can also lead to distortion of the fault signal, making it impossible for the detection system to accurately restore the true fault information, thereby affecting the accuracy of fault detection. Therefore, after constructing the polynomial model, that is, after step 120, in some other embodiments, this application can also compensate for the frequency domain distortion of amplitude and phase of the current distorted signal.

[0075] Step 130: Substitute the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal; where t takes the value of an integer greater than 0 in turn until n is equal to the total number of times in the current distorted signal, and the target restored signal is obtained.

[0076] It should be noted that, since the non-ideal characteristics of the current transformer have a nonlinear effect on the time domain of the measured signal, this application solves for the time-varying coefficients to construct a polynomial model, and then substitutes the data points of each time step in the current distorted signal into the polynomial model in turn to compensate for the time-domain distortion of the data points of each time step in the current distorted signal, so as to obtain the target restored signal through the compensation of time-domain distortion.

[0077] It should be further noted that the non-ideal characteristics of the current transformer not only have a non-linear effect on the time domain, amplitude, and phase of the measured signal, but also have a dynamic non-linear effect on the time domain of the measured signal. This dynamic non-linear effect changes with time and environment, which can also lead to distortion of the fault signal, making it impossible for the detection system to accurately restore the true fault information, thereby affecting the accuracy of fault detection. Therefore, in step 130, in some other embodiments, this application can also perform dynamic compensation for the time domain distortion of the current distorted signal.

[0078] In other embodiments, regarding the determination of the target restored signal, the first and second time-varying coefficients in the polynomial model can be set to 0, and the square of their absolute values ​​can be used as the residual model. The data point at time t in the current distorted signal is substituted into the polynomial model to obtain the data point at time t in the target restored signal. Based on the residual model, the data point at time t in the current distorted signal, and each time-varying coefficient in the polynomial model, each time-varying coefficient in the polynomial model is updated to obtain the updated polynomial model. The updated polynomial model is then used as the polynomial model. Here, t takes integer values ​​greater than 0 in sequence until t equals the total number of times in the current distorted signal, so as to obtain the target restored signal.

[0079] It should be noted that, since each time-varying coefficient in the polynomial model is updated at each time step, and the residual model is based on the polynomial model, correspondingly, each time-varying coefficient in the residual model is also updated along with the update of each time-varying coefficient in the polynomial model.

[0080] In this application, a polynomial model is constructed, and the first and second time-varying coefficients in the polynomial model are set to 0. The square of the absolute value is then used as the residual model. The data point at time t in the current distorted signal is substituted into the polynomial model to obtain the data point at time t in the target restored signal. Based on the residual model, the data point at time t in the current distorted signal, and each time-varying coefficient in the polynomial model, each time-varying coefficient in the polynomial model is updated. The updated polynomial model is then used as the polynomial model again. That is, by dynamically updating the polynomial model, dynamic compensation for time-domain distortion is performed on the data points at each time in the current distorted signal. Through dynamic compensation for time-domain distortion, the target restored signal is obtained. This not only effectively eliminates the nonlinear influence of the transformer on the time domain of the measured signal, but also further eliminates its dynamic nonlinear influence on the time domain of the measured signal, thereby more accurately restoring the true fault information and improving the detection accuracy of hidden faults in overhead lines.

[0081] In some embodiments, the update method for the polynomial model can be implemented by substituting the data point at time t in the current distorted signal into the residual model to obtain the residual at time t. Then, based on the residual at time t, the data point at time t in the current distorted signal, and the nth time-varying coefficient in the polynomial model, the nth time-varying coefficient at time t+1 is determined, where n takes integer values ​​greater than -1 until n equals the total number of time-varying coefficients. This yields the various time-varying coefficients at time t+1. Finally, based on each time-varying coefficient at time t+1, each time-varying coefficient in the polynomial model is updated to obtain the updated polynomial model.

[0082] In this application, by dynamically updating the time-varying coefficients of the polynomial model, dynamic nonlinear compensation in the time domain for the current distorted signal is achieved, which significantly improves the signal restoration accuracy and thus eliminates the influence of the non-ideal characteristics of the transformer on the detection accuracy.

[0083] Furthermore, regarding the determination of the residual at time t, in some embodiments, the expression of the residual model can be used. Obtain the residual at time t; where, Let be the residual at time t. For the (n+1)th time-varying coefficient, The total number of time-varying coefficients. Let t be the data point at time t in the current distorted signal.

[0084] In this embodiment, the residual at time t is determined by using the residual model expression, which provides a key basis for dynamically updating the time-varying coefficients of the polynomial model, effectively improving the signal restoration accuracy and thus eliminating the influence of the non-ideal characteristics of the transformer on the detection accuracy.

[0085] Furthermore, regarding the determination of the nth time-varying coefficient at time t+1, in some embodiments, a formula can be used. Determine the nth time-varying coefficient at time t+1; where, This is the nth time-varying coefficient at time t+1. For the (n+1)th time-varying coefficient, To preset the learning rate, Let be the residual at time t. Let t be the data point at time t in the current distorted signal.

[0086] It should be noted that the preset learning rate can be obtained and set in advance by the operator based on a large amount of experience, experiments or statistics. Of course, it can also be set in advance by the operator according to actual needs.

[0087] In this application, the time-varying coefficients of the polynomial model are adaptively updated through a dynamic formula, which significantly improves the accuracy and efficiency of time-domain dynamic nonlinear compensation.

[0088] In some embodiments, the determination of the data point at time t in the target reconstructed signal can utilize the expression of a polynomial model. The target reconstructed signal is determined (the time-varying coefficients in the polynomial model of this embodiment are updated in real time); wherein, To reconstruct the data point at time t in the target signal, For the (n+1)th time-varying coefficient, The total number of time-varying coefficients. Let t be the data point at time t in the current distorted signal.

[0089] In this application, the data points of the target restored signal at each time moment are accurately determined by the polynomial model expression, which significantly improves the accuracy and reliability of time-domain dynamic nonlinear compensation and eliminates the influence of the non-ideal characteristics of the transformer on the detection accuracy.

[0090] Step 140: Perform multi-feature extraction on the target restored signal to obtain the signal feature vector, and input the signal feature vector into the preset machine learning model to obtain the fault classification and detection result.

[0091] Here, the preset machine learning model refers to a pre-trained model used to predict the output fault classification detection result based on the input signal feature vector.

[0092] For signal feature types extracted from multiple features contained in a signal feature vector, in some embodiments, the signal feature vector includes spectral center frequency, bandwidth, short-time energy, time-domain waveform change, impulse index, kurtosis, margin factor, waveform factor, frequency standard deviation, mean square frequency, energy entropy, etc.

[0093] Regarding the acquisition method of the preset machine learning model, in some embodiments, a large number of experimental reconstruction signals can be acquired, and multi-feature extraction can be performed on the experimental reconstruction signals to obtain experimental signal feature vectors. The labels corresponding to the experimental signal feature vectors can be determined, and then a large number of experimental signal feature vectors and their corresponding labels can be input into the initial machine learning model for training to obtain the preset machine learning model. In other embodiments, a large number of historical distortion signals can also be acquired (i.e., they can be obtained based on experimental reconstruction signals or from line measurements), and the signals can be reconstructed through the above steps 110 to 130 to obtain historical reconstruction signals. Then, multi-feature extraction can be performed on the historical reconstruction signals to obtain historical signal feature vectors, and the labels corresponding to the historical signal feature vectors can be determined. Finally, a large number of historical signal feature vectors and their corresponding labels can be input into the initial machine learning model for training to obtain the preset machine learning model.

[0094] In this embodiment, multiple time-varying coefficients are determined based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimental restored signal to construct a polynomial model. Then, the data point at time t in the current distorted signal is substituted into the polynomial model to obtain the data point at time t in the target restored signal, so as to restore the time domain of the current distorted signal. This method can effectively eliminate the nonlinear influence of the current transformer on the time domain of the measured signal, accurately restore the real fault information, and thus improve the detection accuracy of hidden faults in overhead lines.

[0095] Furthermore, this polynomial model-based line fault detection method, in addition to effectively eliminating the time-domain nonlinear influence of the instrument transformer on the measured signal and improving the accuracy of detecting hidden faults in overhead lines, also has the following advantages: Consideration of amplitude and phase frequency domain distortion compensation: The non-ideal characteristics of the instrument transformer not only have a nonlinear influence on the time domain of the measured signal, but also on the amplitude and phase. After constructing a polynomial model, this method can perform frequency domain distortion compensation for the amplitude and phase of the target restored signal or the current distorted signal, comprehensively eliminating the influence of the instrument transformer's non-ideal characteristics on the signal, making the restored signal closer to the real situation; Realization of time-domain dynamic nonlinear compensation: The non-ideal characteristics of the instrument transformer have a nonlinear influence on the time-domain dynamic nonlinearity of the measured signal. The influence of distortion changes with time and environment. This method dynamically updates the time-varying coefficients of the polynomial model to dynamically compensate for the time-domain distortion of the current distorted signal, further improving the accuracy and reliability of signal restoration and restoring the true fault information with extreme accuracy. The method also features flexible coefficient determination using multi-order polynomials: when constructing a polynomial model with multiple time-varying coefficients, multi-order polynomials are used. The order of these polynomials can be preset by the operator based on extensive experience, experiments, statistics, or actual needs. This flexibility allows the model to better adapt to different scenarios and requirements. Finally, the method uses dynamic formulas to adaptively update the time-varying coefficients of the polynomial model, based on the residuals at each time step and the current distorted signal. The time-varying coefficients in the data points and polynomial model are used to determine the time-varying coefficients at the next moment, significantly improving the accuracy and efficiency of time-domain dynamic nonlinear compensation, enabling the model to adapt to changes in transformer characteristics in a timely manner; the signal generation and acquisition methods are clearly defined: for the generation of experimental reconstruction signals, after obtaining the transformer's operating frequency range, the method of controlling the signal generator to sequentially increase the generated signal frequency within the operating frequency range according to a preset step size to obtain multi-frequency signals is clearly defined, and the operating frequency range is optimized, providing a clear and feasible operation method for acquiring experimental reconstruction signals, which helps to improve the efficiency of the entire fault detection process; the model training data acquisition method is standardized: when acquiring the preset machine learning model, Two methods for acquiring training data are provided: either the experimental signal feature vector can be obtained through experimental signal reconstruction, or the historical signal can be reconstructed from historical distorted signals, and then the historical signal feature vector can be extracted for training. This enriches the sources of training data for the model and helps improve the model's generalization ability and accuracy. Multi-feature extraction enhances feature richness: multi-feature extraction is performed on the target reconstructed signal to obtain signal feature vectors containing various features such as the spectral center frequency, bandwidth, and short-time energy. The rich features can more comprehensively reflect the characteristics of the signal, providing more sufficient information for the preset machine learning model, thereby improving the reliability and stability of fault classification and detection results.Pre-trained machine learning models ensure performance: The pre-trained machine learning models are capable of accurately predicting output fault classification detection results based on the input signal feature vectors. This pre-training method enables the model to quickly and accurately provide detection results in practical applications, improving the efficiency and practicality of fault detection.

[0096] In one feasible implementation, step 130 in the above embodiment, which involves substituting the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal, includes:

[0097] Expression using a polynomial model Obtain the target reconstruction signal;

[0098] in, To reconstruct the data point at time t in the target signal, For the (n+1)th time-varying coefficient, The total number of time-varying coefficients. Let t be the data point at time t in the current distorted signal.

[0099] In this embodiment, the data points of the target reconstructed signal at each moment are accurately determined by a polynomial model expression, which effectively improves the accuracy and reliability of signal reconstruction.

[0100] Understandably, the data point at time t in the target reconstructed signal is determined using a polynomial model expression. This expression includes time-varying coefficients and the corresponding data point in the current distorted signal. Since the time-varying coefficients are accurately calculated based on historical distorted signals and experimental reconstructed signals, they can reflect the characteristic patterns of signal time-domain distortion. By combining these coefficients with the data point at time t in the current distorted signal and performing polynomial operations, the data point of the target reconstructed signal at that time can be accurately calculated. This method fully considers the complex characteristics of signal time-domain distortion, making the determined data point closer to the actual fault signal. This effectively improves the accuracy and reliability of signal reconstruction, providing strong support for the subsequent accurate detection of hidden faults in overhead lines.

[0101] In one feasible implementation, step 120 in the above embodiment, which involves determining multiple time-varying coefficients based on data points at various times in the historical distorted signal and the experimentally restored signal to construct a polynomial model, includes: standardizing the historical distorted signal and the experimentally restored signal to obtain a standard distorted signal and a standard restored signal, respectively; using the least squares method to determine multiple time-varying coefficients based on data points at various times in the standard distorted signal and the standard restored signal; and constructing a polynomial model based on the various time-varying coefficients.

[0102] In this embodiment, the time-varying coefficients of the polynomial model are determined by standardization and the least squares method, which improves the efficiency and accuracy of model construction and thus enhances the effect of time-domain distortion compensation or time-domain distortion dynamic compensation.

[0103] Understandably, standardizing the historical distorted signal and the experimentally reconstructed signal separately yields the standard distorted signal and the standard reconstructed signal. Standardization eliminates dimensional influences and outlier interference, making the signal data more standardized and uniform, providing a more stable data foundation for subsequent model construction. Using the least squares method, multiple time-varying coefficients are determined based on data points at various times in the standard distorted signal and the standard reconstructed signal. The least squares method is a classic mathematical optimization method that finds the best function match for the data by minimizing the sum of squared errors, effectively handling noise and errors in the signal data, making the determined time-varying coefficients more accurate and reliable. A polynomial model is then constructed based on these accurate time-varying coefficients. Because the accuracy of the time-varying coefficients is guaranteed, the constructed polynomial model can more accurately describe the relationship between the historical distorted signal and the experimentally reconstructed signal of the instrument transformer. Therefore, when performing time-domain distortion compensation or dynamic time-domain distortion compensation on the target reconstructed signal, it can more effectively eliminate the nonlinear influence or dynamic nonlinearity of the non-ideal characteristics of the instrument transformer on the signal time domain, improving the accuracy of fault detection.

[0104] In one feasible implementation, the standardization of the historical distorted signal and the experimental restored signal in the above embodiments to obtain a standard distorted signal and a standard restored signal includes: determining a first mean and a first standard deviation of the historical distorted signal, and determining a second mean and a second standard deviation of the experimental restored signal; standardizing the data points at each time point in the historical distorted signal according to the first mean and the first standard deviation to obtain a standard distorted signal, and standardizing the data points at each time point in the experimental restored signal according to the second mean and the second standard deviation to obtain a standard restored signal.

[0105] In the embodiments of this application, the standardization process based on the mean and standard deviation improves the standardization and comparability of the signal data, laying a solid foundation for the subsequent accurate determination of time-varying coefficients and the construction of polynomial models.

[0106] Understandably, determining the first mean and first standard deviation of the historical distorted signal, and the second mean and second standard deviation of the experimentally reconstructed signal, allows us to comprehensively describe the distribution characteristics of the signal data. The mean reflects the central tendency of the data, while the standard deviation reflects the dispersion. Standardizing the data points at each moment in the historical distorted signal based on the first mean and first standard deviation yields the standard distorted signal. Similarly, standardizing the data points at each moment in the experimentally reconstructed signal based on the second mean and second standard deviation yields the standard reconstructed signal. This standardization process unifies signal data with different dimensions and distributions to the same scale, eliminating dimensional and distributional differences and enhancing the data's standardization and comparability. This allows for a more accurate identification of the intrinsic relationships between the data when using the least squares method to determine the time-varying coefficients based on the standard distorted signal and the standard reconstructed signal, avoiding errors caused by inconsistent data scales. This provides a reliable guarantee for constructing a more accurate polynomial model, ultimately improving the effectiveness of time-domain distortion compensation.

[0107] In one feasible implementation, step 120 in the above embodiment, after determining multiple time-varying coefficients based on data points at various times in the historical distorted signal and data points at various times in the experimentally restored signal to construct a polynomial model, further includes: performing Fourier transforms on the current distorted signal, the historical distorted signal, and the experimentally restored signal respectively to obtain the current frequency domain distorted signal, the historical frequency domain distorted signal, and the experimentally restored frequency domain signal; constructing a frequency response transfer function based on the historical frequency domain distorted signal and the experimentally restored frequency domain signal; transforming the frequency response transfer function to obtain a frequency domain form function, and determining the target frequency domain restored signal based on the amplitude attenuation coefficient and phase offset of the frequency domain form function, as well as the amplitude and phase of each frequency in the current frequency domain distorted signal; performing an inverse Fourier transform on the target frequency domain restored signal to obtain an initial restored signal; and using the initial restored signal as the current distorted signal.

[0108] It should be noted that the frequency response transfer function of this application belongs to the s-domain of the Laplace transform, while the frequency form function belongs to the frequency domain of the Fourier transform. Therefore, in some embodiments, the frequency response transfer function can be converted between the s-domain and the frequency domain to obtain the frequency domain form function; where s is the Laplace operator.

[0109] It should be further noted that, since the non-ideal characteristics of the transformer have a non-linear effect on the amplitude and phase of the measured signal, this application compensates for the amplitude and phase of each frequency in the current frequency-domain distorted signal by using the amplitude attenuation coefficient and phase offset of the frequency-domain form function, so as to obtain the target frequency-domain restored signal through the compensation of frequency-domain distortion.

[0110] Furthermore, it should be noted that the current distortion signal initially measured by the current transformer in this application is a time-domain signal, while the target frequency-domain restored signal is a frequency-domain signal. Therefore, in some embodiments, the target frequency-domain restored signal can be subjected to inverse Fourier transform in both the frequency and time domains to obtain the initial restored signal in the time domain.

[0111] Regarding the construction method of the frequency response transfer function, in some embodiments, the frequency response of each frequency can be determined based on the frequency domain components of each frequency in the historical frequency domain distortion signal and the frequency domain components of each frequency in the experimental frequency domain reconstruction signal. Based on the frequency response of each frequency and the preset objective function, multiple numerator transfer coefficients and multiple denominator transfer coefficients are obtained to construct the frequency response transfer function. The preset objective function can be obtained and preset by the operator based on a large amount of experience, experiments or statistics. Of course, it can also be preset by the operator according to actual needs.

[0112] It should be noted that the purpose of the preset objective function is to find the multiple numerator transfer coefficients and multiple denominator transfer coefficients corresponding to each frequency response when the frequency response of each frequency reaches its minimum, so that the frequency response transfer function can be constructed based on the multiple numerator transfer coefficients and multiple denominator transfer coefficients.

[0113] Furthermore, regarding the construction methods of the preset objective function and the frequency response transfer function, in some embodiments, the transfer function can be constructed by having multiple polynomials in both the numerator and denominator. Then, the preset objective function is constructed based on the transfer function and the frequency response at each frequency. The frequency response at each frequency is known, while in the transfer function, only the coefficients of the multiple polynomials in the numerator and denominator are unknown and adjustable. Therefore, by adjusting the coefficients of the multiple polynomials in the numerator and denominator to minimize the frequency response at each frequency, the adjusted coefficients of the multiple polynomials in the numerator are used as numerator transfer coefficients, and the coefficients of the multiple polynomials in the denominator are used as denominator transfer coefficients. Substituting these multiple numerator and denominator transfer coefficients into the transfer function yields the constructed frequency response transfer function.

[0114] Furthermore, regarding the construction method of the frequency response transfer function, in some embodiments, the quotient between the frequency domain component of the d-th frequency in the historical frequency domain distorted signal and the frequency domain component of the d-th frequency in the experimental frequency domain reconstructed signal can be used as the frequency response of the d-th frequency. Here, d takes integer values ​​greater than 0 in sequence until d equals the total number of frequencies in the historical frequency domain distorted signal or the experimental frequency domain reconstructed signal, thus obtaining the frequency response of each frequency. Then, using the LM method, under the condition of satisfying the preset objective function, the transfer coefficients of each numerator and each denominator in the preset objective function are adjusted according to the frequency response of each frequency to minimize the function value of the preset objective function, thus obtaining multiple transfer coefficients of each numerator and multiple transfer coefficients of each denominator. Finally, the frequency response transfer function is constructed based on the transfer coefficients of each numerator and each denominator. Here, LM stands for Levenberg-Marquardt, also known as the Levenberg-Marquardt method.

[0115] In this embodiment, the construction process of the frequency response transfer function is optimized by using the LM method, which significantly improves the convergence speed and accuracy of the transfer coefficient solution. At the same time, it enhances the model's adaptability to complex nonlinear distortions, providing a more reliable foundation for subsequent signal restoration and fault detection.

[0116] Furthermore, the expression for the preset objective function is:

[0117] ;in, In the above formula, The function value of the preset objective function, The total number of frequencies in the historical frequency domain distorted signal or the experimental frequency domain restored signal. Let d be the frequency response at the d-th frequency. Let d be the Laplace operator corresponding to the d-th frequency. The transfer coefficient of the (m+1)th molecule. The total number of numerator or denominator transfer coefficients. The transitivity coefficient of the (m+1)th denominator. The imaginary unit, Pi This is the d-th frequency.

[0118] In this embodiment, by constructing a composite objective function based on frequency response, the fitting accuracy of the transfer function to the non-ideal characteristics of the transformer is significantly improved, while the anti-interference capability of the model in complex electromagnetic environments is enhanced, providing a more reliable mathematical basis for subsequent signal restoration.

[0119] In this embodiment, the frequency response of each frequency is determined based on the frequency components of each frequency in the historical frequency-domain distorted signal and the frequency components of each frequency in the experimental frequency-domain restored signal. Multiple numerator transfer coefficients and multiple denominator transfer coefficients are obtained based on the frequency responses of each frequency and a preset objective function to construct a frequency response transfer function. Then, based on the amplitude attenuation coefficient and phase offset of the frequency domain transfer function corresponding to the constructed frequency response transfer function, the amplitude and phase of the current frequency-domain distorted signal are restored. This not only effectively eliminates the nonlinear effects of the transformer on the measured signal in the time domain, as well as the dynamic nonlinear effects in the time domain, but also further eliminates its nonlinear effects on the amplitude and phase of the measured signal, thereby restoring the true fault information with extreme accuracy and improving the detection accuracy of hidden faults in overhead lines.

[0120] In one feasible implementation, the expression for the frequency response transfer function in the above embodiments is:

[0121] ;

[0122] in, ;

[0123] In the above formula, The frequency response transfer function, For the Laplace operator, The transfer coefficient of the (m+1)th molecule. The total number of numerator or denominator transfer coefficients. The transitivity coefficient of the (m+1)th denominator. The imaginary unit, Pi For frequency.

[0124] In this embodiment, by constructing a frequency response transfer function based on the Laplace domain, high-precision mathematical modeling of the non-ideal characteristics of the transformer is achieved, which significantly improves the amplitude / phase accuracy of signal reconstruction and enhances the model's adaptability to nonlinear distortion, providing a reliable signal basis for subsequent fault feature extraction.

[0125] Understandably, the full-band dynamic compensation capability is achieved by using a multi-order polynomial structure in the transfer function, which can accurately fit the amplitude and phase frequency characteristics of the transformer within the operating frequency range. The complex domain parameter optimization mechanism is achieved by introducing the imaginary unit and the Laplace operator, which allows the transfer function to simultaneously optimize both the real part (amplitude response) and the imaginary part (phase response).

[0126] In one feasible implementation, the expression for the frequency domain form function in the above embodiments is:

[0127] ;

[0128] in, It is a function in the frequency domain. The amplitude attenuation coefficient at frequency f is... It is a natural constant. The imaginary unit, This is the phase offset at frequency f.

[0129] In this embodiment, an independent compensation mechanism for amplitude attenuation and phase shift is implemented by constructing a frequency domain form function, which significantly improves the accuracy and flexibility of frequency domain distortion compensation. At the same time, it provides high-precision initial conditions for subsequent operations and effectively solves the distortion problem caused by the non-ideal characteristics of the transformer.

[0130] Understandably, the amplitude-phase decoupling compensation mechanism decouples the amplitude attenuation coefficient and phase offset into independent parameters in the frequency domain form, allowing for precise compensation for amplitude-frequency and phase-frequency characteristic distortions of the transformer, respectively. The complex domain dynamic modeling capability introduces natural constants and imaginary units, allowing the frequency domain form to be expressed in complex form, thus simultaneously representing the real part (amplitude) and imaginary part (phase) of the signal. This modeling method enables the model to accurately match the nonlinear offset caused by the transformer. The frequency band adaptive compensation characteristic, where both the amplitude attenuation coefficient and phase offset are functions of frequency, dynamically adapts to the differences in nonlinear characteristics of the transformer in different frequency bands.

[0131] In one feasible implementation, determining the target frequency domain reconstructed signal based on the amplitude attenuation coefficient and phase offset of the frequency domain formal function, and the amplitude and phase of each frequency in the current frequency domain distorted signal, in the above embodiments includes:

[0132] Using formula Determine the initial frequency domain reconstructed signal;

[0133] in, To reconstruct the frequency domain component of the target frequency domain signal at the k-th frequency, For the kth frequency, Let be the amplitude of the k-th frequency in the current frequency domain distorted signal. For frequency f k The amplitude attenuation coefficient, It is a natural constant. The imaginary unit, The phase of the k-th frequency in the current frequency domain distorted signal. For frequency f k The phase offset.

[0134] In this embodiment, the amplitude attenuation and phase shift caused by the mutual inductor are accurately and independently compensated by the frequency domain component compensation formula, which significantly improves the restoration accuracy of the frequency domain distortion signal and provides a high-fidelity signal basis for subsequent fault feature extraction.

[0135] Understandably, the amplitude-phase decoupling compensation mechanism decouples the amplitude compensation term from the phase compensation term in the formula, allowing for independent optimization of the amplitude-frequency characteristic distortion and phase-frequency characteristic distortion of the current transformer, respectively. The complex domain dynamic modeling capability introduces natural constants and imaginary units, allowing the formula to be expressed in complex form, thus simultaneously representing the real part (amplitude) and the imaginary part (phase) of the signal. This modeling method enables the model to accurately match the nonlinear offset caused by the current transformer. The frequency band adaptive compensation characteristic is also noteworthy: both the amplitude attenuation coefficient and the phase offset are functions of frequency, allowing for dynamic adaptation to the differences in nonlinear characteristics of the current transformer in different frequency bands.

[0136] In one feasible implementation, the process of performing a Fourier transform on the current distorted signal to obtain the current frequency domain distorted signal in the above embodiments includes: performing a Fourier transform on the current distorted signal to obtain an initial frequency domain distorted signal; determining the maximum frequency based on the experimentally restored signal; determining the cutoff frequency based on the maximum frequency and a preset safety factor; and performing frequency filtering on the initial frequency domain distorted signal based on the cutoff frequency to obtain the current frequency domain distorted signal.

[0137] The preset safety factor can be set in advance by the operator based on a large amount of experience, experiments or statistics. Of course, it can also be set in advance by the operator according to actual needs.

[0138] Regarding the value of the preset safety factor, in some embodiments, this application preferably sets the preset safety factor to 1.2.

[0139] Regarding the method for determining the maximum frequency, in some embodiments, the experimental restored signal can be subjected to Fourier transform to obtain the experimental frequency domain restored signal, and then the highest frequency among the frequency domain components of the experimental frequency domain restored signal can be taken as the maximum frequency; in other embodiments, the operating frequency range of the current transformer can also be used to determine the maximum frequency, such as when the operating frequency range of the current transformer is 0.1Hz to 10kHz, 10kHz can be taken as the maximum frequency.

[0140] Regarding the method for determining the cutoff frequency, in some embodiments, the product of the maximum frequency and the preset safety factor can be used as the cutoff frequency.

[0141] In some embodiments, the frequency components with frequencies greater than the cutoff frequency in the frequency components of the initial frequency-domain distortion signal can be removed to obtain the current frequency-domain distortion signal.

[0142] In this embodiment, the frequency filtering mechanism effectively suppresses high-frequency noise interference, significantly improving the signal-to-noise ratio of the fault signal and the accuracy of subsequent feature extraction.

[0143] Understandably, the adaptive frequency band constraint—determining the maximum frequency based on the experimental frequency domain reconstructed signal and setting the cutoff frequency with a safety factor of 1.2—preserves the effective fault characteristic frequency band while avoiding high-frequency noise contamination of the signal, making it more adaptable than a fixed frequency band scheme; improved anti-aliasing effect—precisely eliminating components outside the cutoff frequency after Fourier transform, effectively preventing high-frequency noise from aliasing into the baseband frequency band, providing a cleaner input signal for subsequent frequency domain distortion compensation; and transformer characteristic adaptation—the cutoff frequency mechanism can automatically match the equipment characteristics to the differences in the operating frequency bands of different transformers, avoiding the characteristic loss or noise introduction problems caused by a fixed frequency band.

[0144] In one feasible implementation, step 140 in the above embodiment, which involves extracting multiple features from the target reconstructed signal to obtain a signal feature vector, includes: performing wavelet transform on the target reconstructed signal to obtain a first time-frequency diagram; performing short-time Fourier transform on the target reconstructed signal to obtain a second time-frequency diagram; performing multiple feature extraction on the first and second time-frequency diagrams to obtain the spectral center frequency, bandwidth, short-time energy, and time-domain waveform change; and using the spectral center frequency, bandwidth, short-time energy, and time-domain waveform change as the signal feature vector.

[0145] In this embodiment, the feature extraction is fused using a multi-time-frequency analysis method, which significantly improves the comprehensiveness of fault signal feature representation and the accuracy of fault classification and detection.

[0146] Understandably, the enhanced time-frequency joint representation combines the complementary characteristics of wavelet transform and short-time Fourier transform, retaining the time-frequency localization analysis capability of wavelet transform while utilizing the frequency resolution advantage of short-time Fourier transform, thus constructing a more complete signal time-frequency feature space. Multi-dimensional feature fusion extracts four core features: spectral center frequency, bandwidth, short-time energy, and time-domain waveform variation, forming a multi-dimensional feature vector covering the frequency, time, and energy domains, effectively solving the problem of insufficient representation of complex fault modes by single features. Improved anti-interference capability enhances the ability of time-frequency analysis methods to handle non-stationary fault signals. Natural adaptability: Time-frequency plot processing can suppress noise interference, and the extracted bandwidth and time-domain waveform changes are more sensitive to hidden faults; Enhanced feature interpretability: All four extracted features have clear physical meaning and are strongly correlated with fault types, providing interpretable input features for machine learning models and helping to improve the credibility of fault classification results; Optimized computational efficiency: A standardized time-frequency transformation process is adopted, and both wavelet transform and short-time Fourier transform can be implemented with fast algorithms. The feature extraction process maintains linear complexity, ensuring feature quality while meeting the requirements of real-time detection.

[0147] In its second aspect, this application provides a line fault detection device based on a polynomial model.

[0148] Please see Figure 2 This is a schematic diagram of a line fault detection device based on a polynomial model in an embodiment of this application. The device 210 includes:

[0149] The acquisition module 211 is used to acquire the current distortion signal and historical distortion signal of the current transformer, as well as the experimental reconstruction signal corresponding to the historical distortion signal;

[0150] The determination and construction module 212 is used to determine multiple time-varying coefficients based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimentally restored signal, so as to construct a polynomial model.

[0151] The data points at time t in the current distorted signal are successively substituted into module 213 to obtain the data points at time t in the target restored signal; where t takes the values ​​of integers greater than 0 until n is equal to the total number of times in the current distorted signal, thus obtaining the target restored signal.

[0152] The extraction and prediction module 214 is used to extract multiple features from the target restored signal to obtain the signal feature vector, and input the signal feature vector into the preset machine learning model to obtain the fault classification detection result.

[0153] In this embodiment of the application, the relevant contents of the above-mentioned acquisition module 211, determination and construction module 212, sequential substitution module 213 and extraction and prediction module 214 can be found in the following references. Figure 1 The contents of the illustrated embodiments will not be repeated here.

[0154] It should be noted that the device 210 of this application also includes other modules. It can be understood that the method of this application and the device 210 have a one-to-one correspondence. Therefore, the other modules of the device 210 of this application are the contents corresponding to the method of this application in the above embodiments.

[0155] In this embodiment, multiple time-varying coefficients are determined based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimental restored signal to construct a polynomial model. Then, the data point at time t in the current distorted signal is substituted into the polynomial model to obtain the data point at time t in the target restored signal, so as to restore the time domain of the current distorted signal. This device can effectively eliminate the nonlinear influence of the current transformer on the time domain of the measured signal, accurately restore the real fault information, and thus improve the detection accuracy of hidden faults in overhead lines.

[0156] In addition to effectively eliminating the time-domain nonlinear effects of instrument transformers on measured signals and improving the accuracy of detecting hidden faults in overhead lines, this polynomial model-based line fault detection device also has the following advantages: It considers amplitude and phase frequency domain distortion compensation: The non-ideal characteristics of instrument transformers not only have a nonlinear effect on the time domain of the measured signal, but also on the amplitude and phase. After constructing a polynomial model, this device can perform frequency domain distortion compensation for the target restored signal or the current distorted signal, comprehensively eliminating the influence of the non-ideal characteristics of instrument transformers on the signal, making the restored signal closer to the real situation; it achieves dynamic nonlinear time-domain compensation: The non-ideal characteristics of instrument transformers have a dynamic nonlinear effect on the time domain of the measured signal. The effects of distortion change over time and with the environment. This device dynamically updates the time-varying coefficients of the polynomial model to compensate for the time-domain distortion of the current distorted signal, further improving the accuracy and reliability of signal restoration and accurately restoring the true fault information. The device also features flexible coefficient determination using multi-order polynomials: when constructing a polynomial model with multiple time-varying coefficients, multi-order polynomials are used, whose order can be preset by the operator based on extensive experience, experiments, statistics, or actual needs. This flexibility allows the model to better adapt to different scenarios and requirements. Finally, the device uses dynamic formulas to adaptively update the time-varying coefficients of the polynomial model, based on the residuals at each time step and the current distorted signal. The time-varying coefficients in the data points and polynomial model are used to determine the time-varying coefficients at the next moment, significantly improving the accuracy and efficiency of time-domain dynamic nonlinear compensation, enabling the model to adapt to changes in transformer characteristics in a timely manner; the signal generation and acquisition methods are clearly defined: for the generation of experimental reconstruction signals, after obtaining the transformer's operating frequency range, a device is defined to control the signal generator to sequentially increase the generated signal frequency within the operating frequency range according to a preset step size to obtain multi-frequency signals, and the operating frequency range is optimized, providing a clear and feasible operation method for acquiring experimental reconstruction signals, which helps improve the efficiency of the entire fault detection process; the model training data acquisition method is standardized: when acquiring the preset machine learning model, Two methods for acquiring training data are provided: either the experimental signal feature vector can be obtained through experimental signal reconstruction, or the historical signal can be reconstructed from historical distorted signals, and then the historical signal feature vector can be extracted for training. This enriches the sources of training data for the model and helps improve the model's generalization ability and accuracy. Multi-feature extraction enhances feature richness: multi-feature extraction is performed on the target reconstructed signal to obtain signal feature vectors containing various features such as the spectral center frequency, bandwidth, and short-time energy. The rich features can more comprehensively reflect the characteristics of the signal, providing more sufficient information for the preset machine learning model, thereby improving the reliability and stability of fault classification and detection results.Pre-trained machine learning models ensure performance: The pre-trained machine learning models are capable of accurately predicting output fault classification detection results based on the input signal feature vectors. This pre-training method enables the model to quickly and accurately provide detection results in practical applications, improving the efficiency and practicality of fault detection.

[0157] In a third aspect, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform a line fault detection method based on a polynomial model as described in any one of the first aspects.

[0158] This application provides a computer device in a fourth aspect, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform a line fault detection method based on a polynomial model as described in any one of the first aspects.

[0159] Figure 3 The diagram illustrates the internal structure of a computer device in some embodiments. This computer device may specifically be a terminal, a server, or a gateway. Figure 3 As shown, the computer device includes a processor, memory, and network interface connected via a system bus.

[0160] The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When executed by a processor, this computer program causes the processor to perform the steps in the above method embodiments. The internal memory may also store a computer program, which, when executed by a processor, causes the processor to perform the steps in the above method embodiments. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0161] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods.

[0162] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0163] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0164] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A line fault detection method based on a polynomial model, characterized in that, The method includes: Acquire the current distortion signal and historical distortion signal of the current transformer, as well as the experimental reconstruction signal corresponding to the historical distortion signal; Based on the data points at each time point in the historical distorted signal and the data points at each time point in the experimentally restored signal, multiple time-varying coefficients are determined to construct a polynomial model. Substitute the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal; where t takes the values ​​of integers greater than 0 in turn until t equals the total number of times in the current distorted signal, thus obtaining the target restored signal; The target restored signal is subjected to multi-feature extraction to obtain a signal feature vector, and the signal feature vector is input into a preset machine learning model to obtain a fault classification and detection result; The step of extracting multiple features from the target restored signal to obtain a signal feature vector includes: Perform wavelet transform on the target restored signal to obtain the first time-frequency diagram; Perform a short-time Fourier transform on the target restored signal to obtain a second time-frequency diagram; Multi-feature extraction is performed on the first time-frequency graph and the second time-frequency graph to obtain the spectrum center frequency, bandwidth, short-time energy and time-domain waveform change. The center frequency of the spectrum, the bandwidth, the short-time energy, and the time-domain waveform change are used as the signal feature vector.

2. The line fault detection method based on a polynomial model according to claim 1, characterized in that, The step of substituting the data point at time t in the current distorted signal into the polynomial model to obtain the data point at time t in the target restored signal includes: Using the expression of the polynomial model The target restored signal is obtained; in, Let be the data point at time t in the target restored signal. For the (n+1)th time-varying coefficient, Subtract 1 from the total number of time-varying coefficients. Let be the data point at time t in the current distorted signal.

3. The line fault detection method based on a polynomial model according to claim 1, characterized in that, The step of determining multiple time-varying coefficients based on data points at various times in the historical distorted signal and data points at various times in the experimentally restored signal to construct the polynomial model includes: The historical distortion signal and the experimental restored signal are standardized respectively to obtain the standard distortion signal and the standard restored signal; Using the least squares method, multiple time-varying coefficients are determined based on the data points at each time step in the standard distorted signal and the data points at each time step in the standard restored signal. The polynomial model is constructed based on each time-varying coefficient.

4. The line fault detection method based on a polynomial model according to claim 3, characterized in that, The standardization process, which involves processing the historical distortion signal and the experimental restored signal to obtain a standard distortion signal and a standard restored signal, includes: Determine the first mean and first standard deviation of the historical distorted signal, and determine the second mean and second standard deviation of the experimentally restored signal; Based on the first mean and the first standard deviation, the data points at each time point in the historical distortion signal are standardized to obtain the standard distortion signal. Based on the second mean and the second standard deviation, the data points at each time point in the experimental restored signal are standardized to obtain the standard restored signal.

5. The line fault detection method based on a polynomial model according to claim 1, characterized in that, After determining multiple time-varying coefficients to construct a polynomial model based on data points at various times in the historical distorted signal and data points at various times in the experimentally reconstructed signal, the method further includes: Fourier transforms are performed on the current distorted signal, the historical distorted signal, and the experimental restored signal, respectively, to obtain the current frequency domain distorted signal, the historical frequency domain distorted signal, and the experimental frequency domain restored signal; Based on the historical frequency domain distortion signal and the experimental frequency domain restored signal, a frequency response transfer function is constructed. The frequency response transfer function is transformed to obtain a frequency domain form function, and the target frequency domain restored signal is determined based on the amplitude attenuation coefficient and phase offset of the frequency domain form function, as well as the amplitude and phase of each frequency in the current frequency domain distorted signal. Perform an inverse Fourier transform on the target frequency domain reconstructed signal to obtain the initial reconstructed signal; The initial restored signal is used as the current distortion signal.

6. The line fault detection method based on a polynomial model according to claim 5, characterized in that, The expression for the frequency response transfer function is: ； in, ; In the above formula, The frequency response transfer function is... For the Laplace operator, The transfer coefficient of the (m+1)th molecule. Subtract 1 from the total number of numerator or denominator transfer coefficients. The transitivity coefficient of the (m+1)th denominator. The imaginary unit, Pi For frequency.

7. The line fault detection method based on a polynomial model according to claim 5, characterized in that, The expression for the frequency domain form function is: ； in, For the frequency domain form function, The amplitude attenuation coefficient at frequency f is... It is a natural constant. The imaginary unit, The phase offset is the frequency f.

8. The line fault detection method based on a polynomial model according to claim 5, characterized in that, The step of determining the target frequency domain reconstructed signal based on the amplitude attenuation coefficient and phase offset of the frequency domain form function, and the amplitude and phase of each frequency in the current frequency domain distorted signal, includes: Using formula Determine the target frequency domain reconstructed signal; in, The frequency domain component of the k-th frequency in the target frequency domain reconstructed signal. For the kth frequency, Let x be the amplitude of the k-th frequency in the current frequency domain distorted signal. For frequency f k The amplitude attenuation coefficient, It is a natural constant. The imaginary unit, The phase of the k-th frequency in the current frequency domain distorted signal. For frequency f k The phase offset.

9. The line fault detection method based on a polynomial model according to claim 5, characterized in that, Performing a Fourier transform on the current distorted signal yields the current frequency domain distorted signal, including: Perform a Fourier transform on the current distorted signal to obtain the initial frequency domain distorted signal; Based on the reconstructed signal from the experiment, determine the maximum frequency; The cutoff frequency is determined based on the maximum frequency and the preset safety factor; Based on the cutoff frequency, the initial frequency domain distortion signal is frequency-filtered to obtain the current frequency domain distortion signal.

Images