A method for detecting fault signals of AC / DC hybrid power lines

Through the CEEMDAN algorithm and support vector machine classifier, combined with HFDE entropy value and fluctuation coefficient, fault characteristic vector is constructed, which solves the problems of noise interference and high-frequency component omission in AC-DC hybrid power line fault monitoring, and improves the accuracy of fault detection and reliability of power systems.

CN118501613BActive Publication Date: 2024-11-05TIANJIN UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410602543.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-15
Publication Date
2024-11-05
Estimated Expiration
2044-05-15

AI Technical Summary

Technical Problem

The existing AC-DC mixed power line fault monitoring methods are difficult to effectively suppress noise interference and miss high-frequency components, resulting in misjudgment or misjudgment, affecting the accuracy and reliability of fault diagnosis.

Method used

The CEEMDAN algorithm is used to perform signal decomposition, and the signal reconstruction is performed by calculating the correlation coefficient between the IMF component and the original signal, and the fault characteristic vector is constructed in combination with the HFDE entropy value and fluctuation coefficient, and fault monitoring and identification are used by the support vector machine classifier.

Benefits of technology

It improves the accuracy and robustness of signal decomposition, enhances the reliability of fault characteristics, improves the accuracy of fault detection and diagnosis, reduces the maintenance cost of the power system, and improves the quality of power services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118501613B_ABST
    Figure CN118501613B_ABST
Patent Text Reader

Abstract

The invention discloses a method for detecting fault signals of an AC / DC hybrid power line, and relates to the technical field of power fault detection. The technical key points of the method are as follows: a normal signal and a fault signal are input into a CEEMDAN algorithm as sample data to be decomposed, the signal is decomposed into IMF components, the correlation coefficient between each IMF component and an original signal is calculated, a correlation coefficient threshold is pre-set, an IMF component whose correlation coefficient is greater than or equal to the correlation coefficient threshold is selected, and the selected IMF component is signal reconstructed to generate a reconstructed signal, the reconstructed signal is divided into a plurality of sub-segments, a signal of each sub-segment is Fourier transformed, an HFDE entropy value of each sub-segment of the reconstructed signal and a fluctuation coefficient of the reconstructed signal are calculated, a fault feature vector is constructed, the fault signal is monitored and identified, the accuracy and adaptability of signal decomposition are improved, and modal aliasing and noise interference are suppressed at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power fault detection, and in particular to a method for detecting fault signals of an AC / DC hybrid power line. Background Art

[0002] As the power system becomes increasingly complex, fault monitoring and location of AC / DC hybrid power lines have become a key issue to ensure the stable operation of the power system. Existing fault monitoring methods are often targeted at specific types of lines and cannot adapt to the complex environment of AC / DC hybrid power lines. Therefore, it is necessary to develop a method that can accurately monitor and locate AC / DC hybrid power line faults.

[0003] In the Chinese invention application with application publication number CN115267605A, a method and system for monitoring fault signals of AC / DC hybrid power lines are disclosed, which belongs to the field of protection and control technology of AC / DC hybrid power systems. The voltage signal of the fault phase terminal is collected; the voltage of the fault phase terminal is subjected to fast Fourier analysis to obtain the voltage amplitude spectrum of the fault phase terminal, and the fundamental wave and each harmonic content are calculated; the voltage time-shift energy distribution is obtained as the fault feature to distinguish the fault nature and identify the arc extinction moment; the time-shift voltage energy ratio is calculated, and when it is detected that the time-shift voltage energy ratio exceeds a predetermined value, the fault is determined to be an instantaneous fault and the arc extinction moment is determined, and a closing command is issued after a reclosing delay; when the detection voltage energy ratio does not exceed the predetermined value, the reclosing is locked.

[0004] In combination with the prior art, the above application still has the following deficiencies:

[0005] 1. In existing signal processing technologies, it is often difficult to completely suppress noise interference. Noise may come from various sources, such as environmental noise, equipment noise, etc. It will have a serious impact on the quality and reliability of the signal. Especially in power line fault signal monitoring, noise interference may mask the real fault signal, leading to misjudgment or missed judgment.

[0006] 2. Information about high-frequency components is often easily missed. High-frequency components often contain important details and features, such as transient processes or high-frequency oscillations of fault signals. If these high-frequency components are ignored, important information may be lost, affecting the accuracy and reliability of fault diagnosis. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] In view of the deficiencies in the prior art, the present invention provides a method for detecting fault signals of an AC / DC hybrid power line. The method comprises the following steps: taking normal signals and fault signals of multiple groups of AC / DC hybrid power lines as sample data to be decomposed, inputting the signals into a CEEMDAN algorithm, performing adaptive signal decomposition, decomposing the signals into IMF components, calculating the correlation coefficients between the respective IMF components and the original signal, presetting a correlation coefficient threshold, selecting the IMF components whose correlation coefficients are greater than or equal to the correlation coefficient threshold, performing signal reconstruction on the selected IMF components, generating a reconstructed signal x(t), dividing the reconstructed signal x(t) with a length of N into K sub-segments with a length of L, performing Fourier transform on the signal of each sub-segment, calculating the HFDE entropy value of each sub-segment of the reconstructed signal, and the fluctuation coefficient of the reconstructed signal, constructing a fault feature vector, monitoring and identifying the fault signal, and solving the problems mentioned in the background technology.

[0009] (II) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for detecting fault signals of AC / DC hybrid power lines, comprising the following steps:

[0011] A signal acquisition device is installed on the AC / DC hybrid power line, and a sensor is used to collect normal signals and fault signals on the AC / DC hybrid power line to establish a sample data set;

[0012] The normal signals and fault signals of multiple groups of AC / DC hybrid power lines are used as sample data to be decomposed and input into the CEEMDAN algorithm to perform adaptive signal decomposition and decompose the signals into intrinsic mode components, i.e., IMF components.

[0013] Calculate the correlation coefficient between each IMF component and the original signal, pre-set the correlation coefficient threshold, select the IMF component whose correlation coefficient is greater than or equal to the correlation coefficient threshold, and reconstruct the selected IMF component to generate a reconstructed signal x(t);

[0014] Divide the reconstructed signal x(t) of length N into K sub-segments of length L, perform Fourier transform on the signal of each sub-segment, and calculate the HFDE entropy value S of each sub-segment of the reconstructed signal k , calculate the average value to get the final HFDE entropy value S;

[0015] According to the frequency F of each sub-segment k and amplitude A k , obtain the frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal, further calculate the reconstructed signal fluctuation coefficient RS, combine the calculated HFDE entropy value S with the reconstructed signal fluctuation coefficient RS to construct the fault feature vector;

[0016] The constructed fault feature vector is used as the input feature, and the fault type of the AC / DC hybrid power line is used as the output label. A training model of the support vector machine multi-fault classifier is built. The trained SVM classifier is used to monitor and identify fault signals, and the alarm system is automatically triggered when an abnormal or fault signal is found.

[0017] Furthermore, Gaussian white noise is added to the signal to be decomposed x to obtain a new signal x i , for the new signal x i Perform EMD decomposition to obtain the first IMF signal

[0018]

[0019] Calculate and remove the first IMF signal The remaining signal r1 after:

[0020]

[0021] Among them, ω i represents the added Gaussian white noise, i represents the ith noise addition, i = 1, 2, ..., n, and n is a positive integer, ε0 represents the adaptive coefficient when calculating the intrinsic mode function, and ε0 is a constant, It is represented as the i-th component generated after the i-th noise addition.

[0022] Furthermore, the mth residual signal r is calculated by the formula m and the m+1th IMF signal Where m = 1, 2, ..., M, and M is a positive integer. The calculation formula is as follows:

[0023]

[0024] Among them, r m-1 is the m-1th residual signal, is the mth IMF signal, E m () is the mth intrinsic mode function generated by the empirical mode decomposition method, ε m is the adaptive coefficient when calculating the mth intrinsic mode function, ω i represents the added Gaussian white noise, i represents the i-th noise addition, i = 1, 2, ..., n, and n is a positive integer;

[0025] Repeat the decomposition of the signal until the remaining signal can no longer be decomposed by EMD.

[0026] Furthermore, the original signal and the IMF components obtained by decomposing the original signal are obtained, and the correlation coefficient δ between each IMF component and the original signal is calculated. m , the calculation formula is as follows:

[0027] δ m =(xP m )*A m

[0028] Among them, x is the original signal, P m is the frequency of the mth IMF component obtained by Hilbert transform, A m is the amplitude of the mth IMF component, and * indicates the convolution operation.

[0029] Furthermore, a correlation coefficient threshold is pre-set, and when the correlation coefficient δ between the IMF component and the original signal m When it is greater than or equal to the correlation coefficient threshold, it is marked;

[0030] Using the marked IMF components, the signal is restored through the reconstruction formula to generate the reconstructed signal x(t). The reconstruction formula is as follows:

[0031] x(t)=Re(∑IMF(t)*e jωt )

[0032] Wherein, x(t) represents the reconstructed signal, IMF(t) represents the selected IMF component, ω represents the angular frequency of the signal, j represents the imaginary unit, Re() represents taking the real part, and t represents time.

[0033] Further, the reconstructed signal x(t) of length N is divided into K sub-segments of length L (K=N / L), and a segmentation point is defined at the beginning of each sub-segment;

[0034] The signal of each sub-segment is subjected to Fourier transform to obtain the spectrum feature, and the spectrum coefficient threshold is preset to control the number of spectrum coefficients in each sub-segment.

[0035] Furthermore, the signal of each sub-segment is reconstructed by retaining a fixed number of spectral coefficients, and the HFDE entropy value S of each sub-segment of the reconstructed signal is calculated. k , the calculation formula is as follows:

[0036] S k =-∑(P*logP)

[0037] Where P is the distribution probability of the spectral coefficients retained within a fixed number, and k represents the kth subsegment;

[0038] The HFDE entropy value S of all sub-segments k The final HFDE entropy value S is obtained by summarizing and averaging.

[0039] Further, obtain the frequency F of each sub-segment k and amplitude A k, according to the average frequency of all sub - segments and the average amplitude The frequency fluctuation index FV and the amplitude fluctuation index AS of the reconstructed signal are obtained, and the calculation formulas are as follows:

[0040]

[0041] Where K represents the number of sub - segments, and k represents the k - th sub - segment.

[0042] Furthermore, through the frequency fluctuation index FV and the amplitude fluctuation index AS of the reconstructed signal, the fluctuation coefficient RS of the reconstructed signal is obtained, and the calculation formula is as follows:

[0043]

[0044] Where k1 and k2 represent weight coefficients, 0 < k1 < 1, 0 < k2 < 1, and k1 2 + k2 2 = 1;

[0045] The calculated HFDE entropy value S and the fluctuation coefficient RS of the reconstructed signal are combined to construct a fault feature vector.

[0046] Furthermore, taking the constructed fault feature vector as the input feature and the fault types of the AC - DC hybrid power line as the output label, a training model of a support vector machine multi - fault classifier is constructed;

[0047] The trained SVM classifier is used to monitor and identify the fault signal. Taking the input fault feature vector as the input, the result of the fault type is obtained through the prediction output of the SVM classifier;

[0048] For each input fault signal, the SVM classifier conducts real - time monitoring and early warning, and automatically triggers the alarm system when abnormal or fault signals are found.

[0049] (III) Beneficial Effects

[0050] The present invention provides a method for detecting fault signals of an AC - DC hybrid power line, having the following beneficial effects:

[0051] (1) By using the CEEMDAN algorithm to decompose the normal signal and the fault signal, based on the empirical mode decomposition (EMD), through adding noise and ensemble strategies, the IMF is improved to obtain more accurate decomposition results, which can better adapt to the characteristics of voltage and current signals in the AC - DC hybrid power line, improve the accuracy and self - adaptability of signal decomposition, and at the same time suppress mode mixing and noise interference.

[0052] (2) By calculating the correlation coefficient between each IMF component and the original signal, selecting the IMF component with a correlation coefficient greater than or equal to the correlation coefficient threshold, and restoring the signal through the reconstruction formula, it is possible to effectively suppress noise interference, retain signal characteristics, achieve accurate reconstruction, and enhance robustness, thereby improving the effect and reliability of signal processing.

[0053] (3) By dividing the reconstructed signal into multiple sub-segments and calculating the HFDE entropy value of each sub-segment, the complexity and distribution of the signal can be analyzed more comprehensively, thereby detecting faults more accurately. Since the HFDE entropy value is an analysis method for nonlinear and non-stationary signals, it can better capture and extract the fault features in the reconstructed signal and avoid missing information about high-frequency components, thereby enhancing the robustness of the fault features.

[0054] (4) The HFDE entropy value S and the reconstructed signal fluctuation coefficient RS describe the characteristics of the signal from different perspectives. The HFDE entropy value reflects the complexity and irregularity of the signal, while the reconstructed signal fluctuation coefficient emphasizes the fluctuation and change of the signal. Combining the two can more comprehensively describe the characteristics of the fault and improve the accuracy of fault detection and diagnosis.

[0055] (5) By using the constructed fault feature vector as input feature and using the support vector machine multi-fault classifier to monitor and identify the fault signals of AC / DC hybrid power lines, the reliability and safety of the power system can be improved, the maintenance cost can be reduced, the impact of faults on the power system can be avoided, and the quality of power services can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of the steps of a hybrid power line fault signal detection method according to the present invention;

[0057] Figure 2 The figure is a flow chart of a method for detecting a fault signal of a hybrid power line according to the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] See also Figure 1 to Figure 2 The present invention provides a method for detecting fault signals of AC / DC hybrid power lines, comprising the following steps:

[0060] Step 1: Install a signal acquisition device on the AC / DC hybrid power line, use sensors to collect normal signals and fault signals on the AC / DC hybrid power line, and establish a sample data set;

[0061] The step 1 includes the following contents:

[0062] Step 101: Install a signal acquisition device on an AC / DC hybrid power line, including a current sensor and a voltage sensor;

[0063] Step 102: using sensors to collect normal signals and fault signals on the AC / DC hybrid power line;

[0064] Step 103: Divide the collected signals into training samples and test samples to establish a sample data set.

[0065] It should be noted that the acquisition of fault signals of AC / DC hybrid power lines needs to take into account the characteristics and differences of AC and DC signals, select appropriate sensors and acquisition equipment, and perform accurate measurements and data processing. At the same time, it is also necessary to adopt corresponding monitoring strategies and protection measures according to actual conditions to ensure the safe and stable operation of the power system.

[0066] When using, combine the contents of step 101 to step 103:

[0067] By collecting voltage and current signals in real time, the real-time operating status of power lines can be monitored, and faults can be discovered and located in a timely manner, which helps to improve the reliability and stability of the power system and reduce the probability of faults. The collected and stored voltage and current signals can be used for subsequent data analysis, providing a basis for fault signal detection in the power system.

[0068] Step 2: Input the normal signals and fault signals of multiple groups of AC / DC hybrid power lines as sample data to be decomposed into the CEEMDAN algorithm, perform adaptive signal decomposition, and decompose the signals into intrinsic mode components, i.e., IMF components;

[0069] The step 2 includes the following contents:

[0070] Step 201: collecting normal signals and fault signals of multiple groups of AC / DC hybrid power lines as sample data to be decomposed;

[0071] Step 202: performing pre-processing operations such as denoising and filtering on the sample data to improve the accuracy and credibility of the data;

[0072] Step 203: input the preprocessed sample data into the CEEMDAN algorithm to perform adaptive signal decomposition, and decompose the signal into intrinsic mode components (IMFs), namely, IMF components.

[0073] It should be noted that the CEEMDAN algorithm is an adaptive signal decomposition method, which is called adaptive noise complete ensemble empirical mode decomposition in Chinese. It can process nonlinear and non-stationary signals. The algorithm first performs empirical mode decomposition (EMD) on the signal to obtain a series of intrinsic mode components (IMFs). Then, by adding noise and ensemble strategies, the IMF is improved to obtain more accurate decomposition results.

[0074] The decomposition steps of the CEEMDAN algorithm include:

[0075] Step 2031: Add Gaussian white noise to the signal to be decomposed x to obtain a new signal x i , for the new signal x i Perform EMD decomposition to obtain the first IMF signal

[0076]

[0077] Among them, ω i represents the added Gaussian white noise, i represents the ith noise addition, i = 1, 2, ..., n, and n is a positive integer, ε0 represents the adaptive coefficient when calculating the intrinsic mode function, and ε0 is a constant, It is represented as the i-th component generated after the i-th noise addition;

[0078] Step 2032: Calculate and remove the first IMF signal The remaining signal r1 after:

[0079]

[0080] Step 2033: The original signal obtained by EMD processing for each calculation decomposition is:

[0081] r1+ε1E1(ω i )

[0082] Then, the second IMF signal can be obtained from this for:

[0083]

[0084] Where E1() is the intrinsic mode function generated by the empirical mode decomposition method, r1 is the remaining signal after removing the first IMF signal, ε1 is the adaptive coefficient when calculating the intrinsic mode function, ω i represents the added Gaussian white noise, i represents the i-th noise addition, i = 1, 2, ..., n, and n is a positive integer;

[0085] Step 2034: Calculate the mth residual signal r by the formula m and the m+1th IMF signal Where m = 1, 2, ..., M, and M is a positive integer. The calculation formula is as follows:

[0086]

[0087] Among them, r m-1 is the m-1th residual signal, is the mth IMF signal, E m () is the mth intrinsic mode function generated by the empirical mode decomposition method, ε m is the adaptive coefficient when calculating the mth intrinsic mode function, ω i represents the added Gaussian white noise, i represents the i-th noise addition, i = 1, 2, ..., n, and n is a positive integer;

[0088] Step 2035: Repeat the decomposition of the signal until the remaining signal cannot be further decomposed by EMD.

[0089] It should be noted that the definition of the adaptive coefficient is determined based on the relative size between the noise intensity and the signal intensity. Specifically, the adaptive coefficient is the ratio of the noise intensity to the signal intensity, which is used to adjust the impact of noise on signal decomposition. The selection of the adaptive coefficient needs to be adjusted according to the specific application scenario. An excessively large adaptive coefficient will cause excessive noise interference and affect the decomposition accuracy; an excessively small adaptive coefficient may not be able to fully play the role of the adaptive noise method. Therefore, when applying the CEEMDAN algorithm, it is necessary to make appropriate adjustments based on the characteristics of the actual signal and the decomposition requirements.

[0090] When used, combine the contents of step 201 to step 203:

[0091] The normal signal and fault signal are decomposed by the CEEMDAN algorithm. Based on the empirical mode decomposition (EMD), the IMF is improved by adding noise and set strategy to obtain more accurate decomposition results. It can better adapt to the characteristics of voltage and current signals in AC / DC hybrid power lines, improve the accuracy and adaptability of signal decomposition, and suppress modal aliasing and noise interference.

[0092] Step 3: Calculate the correlation coefficient between each IMF component and the original signal, pre-set the correlation coefficient threshold, select the IMF component whose correlation coefficient is greater than or equal to the correlation coefficient threshold, and reconstruct the selected IMF component to generate the reconstructed signal x(t);

[0093] The step three includes the following contents:

[0094] Step 301: Obtain the original signal and the IMF components obtained by decomposing the original signal, and calculate the correlation coefficient δ between each IMF component and the original signalm , the calculation formula is as follows:

[0095] δ m =(xP m )*A m

[0096] Among them, x is the original signal, P m is the frequency of the mth IMF component obtained by Hilbert transform, A m is the amplitude of the mth IMF component, * represents the convolution operation;

[0097] Step 302: Preset the correlation coefficient threshold. When the correlation coefficient δ between the IMF component and the original signal m When it is greater than or equal to the correlation coefficient threshold, it is marked;

[0098] Step 303: Using the marked IMF components, the signal is restored through a reconstruction formula to generate a reconstructed signal x(t). The reconstruction formula is as follows:

[0099] x(t)=Re(∑IMF(t)*e jωt )

[0100] Wherein, x(t) represents the reconstructed signal, IMF(t) represents the selected IMF component, ω represents the angular frequency of the signal, j represents the imaginary unit, Re() represents taking the real part, and t represents time.

[0101] It should be noted that in the processing of AC / DC hybrid power line fault signals, the use of CEEMDAN decomposition and Hilbert transform can extract the characteristic information in the fault signal and obtain the analytical signal corresponding to each IMF component. The instantaneous characteristics of the analytical signal can be used to reconstruct the signal, thereby realizing the recovery or reconstruction of the fault signal.

[0102] When used, combine the contents of step 301 to step 303:

[0103] By calculating the correlation coefficient between each IMF component and the original signal, selecting the IMF component with a correlation coefficient greater than or equal to the correlation coefficient threshold, and restoring the signal through the reconstruction formula, it is possible to effectively suppress noise interference, retain signal characteristics, achieve accurate reconstruction and enhance robustness, thereby improving the effect and reliability of signal processing.

[0104] Step 4: Divide the reconstructed signal x(t) of length N into K sub-segments of length L, perform Fourier transform on the signal of each sub-segment, and calculate the HFDE entropy value S of each sub-segment of the reconstructed signal k , calculate the average value to get the final HFDE entropy value S;

[0105] The step 4 includes the following contents:

[0106] Step 401: Divide the reconstructed signal x(t) of length N into K sub-segments of length L (k=N / L), and define a segmentation point at the beginning of each sub-segment;

[0107] Step 402: Perform Fourier transform on the signal of each sub-segment to obtain spectrum characteristics, and pre-set spectrum coefficient thresholds to control the number of spectrum coefficients in each sub-segment;

[0108] Step 403: reconstruct the signal of each sub-segment by retaining a fixed number of spectrum coefficients, and calculate the HFDE entropy value S of each sub-segment of the reconstructed signal. k , the calculation formula is as follows:

[0109] S k =-∑(P·logP)

[0110] Where P is the distribution probability of the spectral coefficients retained within a fixed number, and k represents the kth subsegment;

[0111] Step 404: Calculate the HFDE entropy values ​​S of all sub-segments k The final HFDE entropy value S is obtained by summarizing and averaging.

[0112] It should be noted that segmented processing and Fourier transform are commonly used techniques in signal processing, which can effectively convert time domain signals into frequency domain representations, thereby extracting fault-related features. The HFDE entropy value is an indicator used to analyze signal complexity, which can reflect the distribution of the signal in the high-frequency part. Combined with other known fault feature quantities, it can more comprehensively reflect the state and characteristics of the fault. When constructing the fault feature vector, it is necessary to pay attention to selecting appropriate features and appropriate processing methods to achieve accurate fault diagnosis.

[0113] When used, combine the contents of step 401 to step 404:

[0114] By dividing the reconstructed signal into multiple sub-segments and calculating the HFDE entropy value of each sub-segment, the complexity and distribution of the signal can be analyzed more comprehensively, thereby detecting faults more accurately. Since the HFDE entropy value is an analysis method for nonlinear and non-stationary signals, it can better capture and extract fault features in the reconstructed signal, avoid missing information about high-frequency components, and thus enhance the robustness of the fault features.

[0115] Step 5: According to the frequency F of each sub-segment k and amplitude A k, obtain the frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal, further calculate the fluctuation coefficient RS of the reconstructed signal, and combine the calculated HFDE entropy value S with the fluctuation coefficient RS of the reconstructed signal to construct a fault feature vector;

[0116] Step 5 includes the following contents:

[0117] Step 501: Obtain the frequency F k and amplitude A k of each sub-segment, and obtain the frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal according to the average frequency and average amplitude of all sub-segments. The calculation formulas are as follows:

[0118]

[0119] where K represents the number of sub-segments, and k represents the k-th sub-segment;

[0120] Step 502: Obtain the fluctuation coefficient RS of the reconstructed signal through the frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal. The calculation formula is as follows:

[0121]

[0122] where k1 and k2 represent weight coefficients, 0 < k1 < 1, 0 < k2 < 1, and k1 2 + k2 2 = 1;

[0123] Step 503: Combine the calculated HFDE entropy value S with the fluctuation coefficient RS of the reconstructed signal to construct a fault feature vector.

[0124] When in use, combine the contents of Step 501 to Step 503:

[0125] The HFDE entropy value S and the fluctuation coefficient RS of the reconstructed signal describe the characteristics of the signal from different perspectives. The HFDE entropy value reflects the complexity and irregularity of the signal, while the fluctuation coefficient of the reconstructed signal emphasizes the fluctuation and change of the signal. Combining the two can more comprehensively describe the characteristics of the fault and improve the accuracy of fault detection and diagnosis.

[0126] Step 6: Use the constructed fault feature vector as the input feature and the fault type of the AC-DC hybrid power line as the output label to construct a training model of a support vector machine multi-fault classifier, and use the trained SVM classifier to monitor and identify the fault signal, and automatically trigger the alarm system when abnormal or fault signals are found.

[0127] Step 6 includes the following contents:

[0128] Step 601: using the constructed fault feature vector as an input feature and the fault type of the AC / DC hybrid power line as an output label to construct a training model of a support vector machine multi-fault classifier;

[0129] Step 602: using the trained SVM classifier to monitor and identify the fault signal, taking the input fault feature vector as input, and obtaining the result of the fault type through the prediction output of the SVM classifier;

[0130] Step 603: For each input fault signal, the SVM classifier performs real-time monitoring and early warning. When an abnormal or fault signal is found, the alarm system is automatically triggered to notify relevant personnel to handle it.

[0131] It should be noted that both the SVM classifier training process and the fault identification process require sufficient sample data and appropriate feature selection. If the sample data is insufficient or the feature selection is inappropriate, the training effect of the SVM classifier and the accuracy of fault identification may be affected. Therefore, in practical applications, it is necessary to select appropriate features and sample data according to the actual situation, and adjust and optimize the parameters of the SVM classifier.

[0132] When used, combine the contents of step 601 to step 603:

[0133] By taking the constructed fault feature vector as input feature and using the support vector machine multi-fault classifier to monitor and identify the fault signals of AC / DC hybrid power lines, the reliability and safety of the power system can be improved, the maintenance cost can be reduced, the impact of faults on the power system can be avoided, and the quality of power service can be improved.

[0134] In the application, the several formulas involved are all calculated by removing dimensions and taking their numerical values. The formula is a formula of the most recent real situation obtained by collecting a large amount of data and performing software simulation. The coefficients in the formula are set by technicians in this field according to actual conditions.

[0135] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware or in combination with computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0136] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0137] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for detecting fault signals of AC / DC hybrid power lines, characterized in that: The following steps are involved: A signal acquisition device is installed on the AC / DC hybrid power line, and a sensor is used to collect normal signals and fault signals on the AC / DC hybrid power line to establish a sample data set; The normal signals and fault signals of multiple groups of AC / DC hybrid power lines are used as sample data to be decomposed and input into the CEEMDAN algorithm to perform adaptive signal decomposition and decompose the signals into intrinsic mode components, i.e., IMF components. Calculate the correlation coefficient between each IMF component and the original signal, select the IMF component whose correlation coefficient is greater than or equal to the correlation coefficient threshold, and reconstruct the selected IMF component to generate a reconstructed signal x(t); Divide the reconstructed signal x(t) of length N into K sub-segments of length L, perform Fourier transform on the signal of each sub-segment, and calculate the HFDE entropy value S of each sub-segment of the reconstructed signal k , calculate the average value to get the final HFDE entropy value S; According to the frequency F of each sub-segment k and amplitude A k , obtain the frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal, further calculate the reconstructed signal fluctuation coefficient RS, combine the calculated HFDE entropy value S with the reconstructed signal fluctuation coefficient RS to construct the fault feature vector; The constructed fault feature vector is used as the input feature, and the fault type of the AC / DC hybrid power line is used as the output label. A training model of the support vector machine multi-fault classifier is built. The trained SVM classifier is used to monitor and identify fault signals, and the alarm system is automatically triggered when an abnormal or fault signal is found.

2. The method for detecting fault signals of AC / DC hybrid power lines according to claim 1, characterized in that: Add Gaussian white noise to the signal to be decomposed x to obtain the new signal x i , for the new signal x i Perform EMD decomposition to obtain the first IMF signal Calculate and remove the first IMF signal The remaining signal r1 after: Among them, ω i represents the added Gaussian white noise, i represents the ith noise addition, i = 1, 2, ..., n, and n is a positive integer, ε0 represents the adaptive coefficient when calculating the intrinsic mode function, and ε0 is a constant, It is represented as the i-th component generated after the i-th noise addition.

3. A method for detecting fault signals of AC / DC hybrid power lines according to claim 2, characterized in that: The mth residual signal r is calculated by the formula m and the m+1th IMF signal Where m = 1, 2, ..., M, and M is a positive integer. The calculation formula is as follows: Among them, r m-1 is the m-1th residual signal, is the mth IMF signal, E m () is the mth intrinsic mode function generated by the empirical mode decomposition method, ε m is the adaptive coefficient when calculating the mth intrinsic mode function, ω i represents the added Gaussian white noise, i represents the i-th noise addition, i = 1, 2, ..., n, and n is a positive integer; Repeat the decomposition of the signal until the remaining signal can no longer be decomposed by EMD.

4. The method for detecting fault signals of AC / DC hybrid power lines according to claim 3, characterized in that: Get the original signal and the IMF components obtained by decomposing the original signal, and calculate the correlation coefficient δ between each IMF component and the original signal m , the calculation formula is as follows: d m =(xP m )*A m Among them, x is the original signal, P m is the frequency of the mth IMF component obtained by Hilbert transform, A m is the amplitude of the mth IMF component, and * indicates the convolution operation.

5. The method for detecting fault signals of AC / DC hybrid power lines according to claim 4, characterized in that: When the correlation coefficient δ between the IMF component and the original signal m When it is greater than or equal to the correlation coefficient threshold, it is marked; Using the marked IMF components, the signal is restored through the reconstruction formula to generate the reconstructed signal x(t). The reconstruction formula is as follows: x(t)=Re(∑IMF(t)*e jωt ) Wherein, x(t) represents the reconstructed signal, IMF(t) represents the selected IMF component, ω represents the angular frequency of the signal, j represents the imaginary unit, Re() represents taking the real part, and t represents time.

6. The method for detecting fault signals of AC / DC hybrid power lines according to claim 1, characterized in that: Divide the reconstructed signal x(t) of length N into K sub-segments of length L, and define a segmentation point at the beginning of each sub-segment; The signal of each sub-segment is subjected to Fourier transform to obtain the spectrum feature, and the spectrum coefficient threshold is preset to control the number of spectrum coefficients in each sub-segment.

7. A method for detecting fault signals of AC / DC hybrid power lines according to claim 6, characterized in that: The signal of each sub-segment is reconstructed by retaining a fixed number of spectral coefficients, and the HFDE entropy value S of each sub-segment of the reconstructed signal is calculated. k , the calculation formula is as follows: S k =-∑(P*log P) Where P is the distribution probability of the spectral coefficients retained within a fixed number, and k represents the kth subsegment; The HFDE entropy value S of all sub-segments k The final HFDE entropy value S is obtained by summarizing and averaging.

8. The method for detecting fault signals of AC / DC hybrid power lines according to claim 1, characterized in that: Get the frequency F of each sub-segment k and amplitude A k , based on the average frequency of all sub-segments and the average amplitude The frequency fluctuation index FV and amplitude fluctuation index AS of the reconstructed signal are obtained, and the calculation formula is as follows: Wherein, K represents the number of sub-segments, and k represents the kth sub-segment.

9. A method for detecting fault signals of AC / DC hybrid power lines according to claim 8, characterized in that: The reconstructed signal fluctuation coefficient RS is obtained by reconstructing the frequency fluctuation index FV and amplitude fluctuation index AS of the signal. The calculation formula is as follows: Among them, k1 and k2 represent weight coefficients, where 0 < k1 < 1, 0 < k2 < 1, and k1 2 + k2 2 = 1; The calculated HFDE entropy value S is combined with the reconstructed signal fluctuation coefficient RS to construct the fault feature vector.

10. The method for detecting fault signals of AC / DC hybrid power lines according to claim 9, characterized in that: The constructed fault feature vector is used as input feature, the fault type of AC / DC hybrid power line is used as output label, and a training model of support vector machine multi-fault classifier is constructed. The trained SVM classifier is used to monitor and identify fault signals. The input fault feature vector is used as input, and the fault type result is obtained through the prediction output of the SVM classifier. For each input fault signal, the SVM classifier performs real-time monitoring and early warning, and automatically triggers the alarm system when an abnormal or fault signal is found.

Citation Information

Patent Citations

  • Fault signal monitoring method and system for alternating-current and direct-current hybrid power line

    CN115267605A

  • Rotor operation state monitoring method based on CEEMDAN

    CN109253872A

  • Rolling bearing fault diagnosis method based on vibration signal

    CN113780056A

  • Ultra-high voltage direct current transmission line fault identification method

    CN116738213A

  • DC fault arc diagnosis method based on ICEEMDAN-KPCA feature extraction and SOA-KELM

    CN117471252A