A fault recognition and positioning method based on wavelet transform and TKEO

Abstract

The application discloses a double-end traveling wave fault location method based on wavelet transform and Teager-Kaiser energy operator (TKEO), which is applied to a 800V direct current hydrogen production line. In order to improve the accuracy of fault identification and the precision of fault location, firstly, wavelet transform is performed on a fault signal, a soft threshold is used for denoising and the signal is reconstructed; characteristic information of each wavelet decomposition layer is extracted and analyzed, and the fault type is determined according to the energy ratio of high-frequency and low-frequency decomposition layers; secondly, TKEO is used to extract the instantaneous energy spectrum after wavelet decomposition, and the sampling points of the first wave head reaching the two ends of the direct current line are accurately calibrated; finally, the double-end ranging method is used to accurately solve the fault distance.

Description

Technical Field

[0001] This invention belongs to the field of power transmission system fault detection technology, specifically relating to a method for identifying and locating faults in DC transmission lines. Background Technology

[0002] Hydrogen energy, as a zero-carbon emission secondary energy source, is gradually becoming an important force driving the energy revolution due to its clean and efficient characteristics. Currently, hydrogen production is mostly achieved through direct current (DC) systems via water electrolysis. However, a fault in a DC line can generate a very large fault current in a short period of time. Lightning strikes can cause high-frequency transient voltages and currents in DC lines, which can seriously affect the safety and stability of DC microgrids. Therefore, protection devices need to accurately identify and locate the fault within a very short time after it occurs.

[0003] Therefore, accurately identifying faults and pinpointing fault locations in a short time is a significant challenge for photovoltaic DC hydrogen production systems.

[0004] Currently, DC system fault identification methods mainly include voltage-current analysis, fault transient energy analysis, and fault traveling wave analysis; fault location methods include traveling wave method, impedance method, fault analysis method, and natural frequency method. Wavefront extraction is particularly crucial in the traveling wave method, and it is commonly implemented using wavelet transform and Hilbert-Huang transform. However, existing methods have many limitations: wavelet entropy and singular entropy are easily affected by noise; high-frequency signal processing is computationally complex and training efficiency is low when combining variational modal decomposition (VMD) with convolutional neural networks (CNN); VMD parameter optimization depends on initial settings and is prone to getting trapped in local optima; Boltzmann machine learning (BML) has weak generalization ability for unknown faults; S-transform and energy operator performance degrades under strong noise; current characteristic method is insensitive to weak faults; methods such as gated recurrent unit (GRU) are sensitive to parameter settings; empirical mode decomposition (EMD) and VMD also suffer from mode aliasing problems; single-ended measurement methods have insufficient accuracy in locating weak or long-distance faults. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention proposes a two-end traveling wave fault location method based on wavelet transform and the Teager-Kaiser energy operator. First, an appropriate number of wavelet decomposition layers is selected to decompose the fault components after phase-mode transform. Second, the fault is identified and classified by calculating the energy ratio of the high-frequency to low-frequency decomposition layers of the zero-mode component. Finally, the linear mode component is processed by TKEO to obtain fault characteristic components with significant waveform abrupt changes, accurately calibrating the initial wavefronts arriving at both detection points, thus achieving high-precision two-end traveling wave fault location.

[0006] Due to the complexity and randomness of lightning discharge, the actual lightning current waveform is extremely irregular. Through research and practice, experts have proposed various mathematical models to approximate its waveform. According to national standards, the lightning current adopts a negative polarity double exponential waveform, as shown in the following formula: In the formula: This represents the instantaneous current value during a lightning strike. denoted as the peak value of the lightning current; k is the correction coefficient; α and β are the attenuation coefficients of the lightning current's tail and wavefront, respectively. The lightning current waveform is as follows: Figure 1 As shown.

[0007] Modulus analysis, as a decoupling analysis method, plays an important role in DC line fault detection. It can effectively decouple complex signals, extract fault features, perform spectrum analysis and dynamic response analysis, thereby improving the accuracy and sensitivity of fault detection and achieving accurate fault location.

[0008] The phase mode transformation matrix used is: In the formula: , These are the zero-mode and line-mode current components after transformation, respectively; , These represent the currents at the positive and negative terminals of a DC line, respectively.

[0009] To quickly and accurately identify the fault type, wavelet decomposition is performed on the zero-mode current component after phase-mode transformation. During decomposition, appropriate wavelet basis functions and the number of wavelet decomposition levels need to be selected. Since the Daubechies wavelet provides good time-frequency localization characteristics through its different lengths and orders, it is suitable for detecting fault signals with complex characteristics, especially when processing high-frequency noise. Considering smoothness and symmetry, db4 is chosen as the basis function for wavelet decomposition, which can effectively preserve the main features of the signal during processing. Different wavelet decomposition levels affect the extraction of signal features. Lower decomposition levels may fail to capture detailed features in the signal, while higher decomposition levels may over-decompose the signal, leading to feature aliasing. To analyze more high-frequency components, the lowest frequency detail should include the frequency band below 1kHz (i.e., the upper limit of the Nth level frequency). The sampling frequency must be less than 1kHz (where f is the sampling frequency of the fault signal). Furthermore, the frequency must be avoided within the 100Hz range (i.e., the lower limit of the Nth layer frequency). A second harmonic (Hz greater than 100Hz) caused by an external fault appears. To reduce computational complexity and quickly decompose the fault signal, the minimum value of N that meets the conditions should be selected. Therefore, the maximum number of decomposition levels can be calculated according to the following formula. value. In the formula: round represents the rounding function.

[0010] Figure 2 This is a comparison of the spectral analysis of different decomposition levels after wavelet decomposition of positive grounding fault, lightning interference, and lightning fault. Figure 2 Spectral analysis of different decomposition layers shows that the amplitude fluctuation of positive grounding fault is mainly concentrated in D4-D11, the amplitude fluctuation of lightning interference is mainly concentrated in D3-D5, and the lightning fault combines the characteristics of the first two faults, with the amplitude fluctuation mainly concentrated in D5-D9.

[0011] A positive grounding fault forms a fault loop with ground, causing severe attenuation of the transient signal during propagation. Figure 3 shows a schematic diagram of the zero-mode component waveform of a positive grounding fault. As can be seen from Figure 3(b), the amplitude of each frequency band gradually decreases with increasing frequency. In the frequency range of 0–5 kHz, after wavelet decomposition, the amplitude of the zero-mode component decreases rapidly, and the amplitude difference between different frequencies is more significant compared to other frequency bands. Between 5–50 kHz, the amplitude remains essentially unchanged.

[0012] Because lightning interference is short-lived and has a small transient current, its transient energy gradually decreases until it disappears. Therefore, the protection device must ensure that it does not malfunction. Lightning interference can be considered as an independent current source superimposed on the transmission line, and its zero-mode component of fault current is shown in Figure 4. As can be seen from Figure 4(b), the amplitude fluctuation is mainly concentrated in the low-frequency range of 0~30kHz, while the amplitude fluctuation in the 30~50kHz frequency band is less. In addition, in the frequency range below 20kHz, the amplitude differences of each frequency component are large, and the steepness of the amplitude change is high. It can be seen that the time-domain oscillation waveform caused by lightning interference leads to oscillations in the amplitude of each frequency in the spectrum analysis.

[0013] When the lightning current amplitude is large, it can easily cause the insulation material to break down, thus forming an electric arc in a short time, damaging electrical equipment and causing a lightning fault. The zero-mode component of a lightning fault is shown in Figure 5. Its spectral characteristics are similar to those of lightning interference in Figure 4, but there are also differences. Since the amplitude fluctuations of a lightning fault are mainly concentrated in the low-frequency band, its spectral characteristics show the superposition characteristics of positive grounding fault and lightning interference. As shown in Figure 5(b), most of the amplitude fluctuations are concentrated in the 0~20kHz range, and the amplitude fluctuations of the frequency components are obvious. After 20kHz, the amplitude fluctuations gradually decrease and tend to stabilize. Although lightning faults contain more harmonic components in the 20~50kHz frequency band, the amplitude attenuation rate in these frequency bands is smaller than that of lightning interference.

[0014] The accuracy of traveling wave signal detection is a key factor affecting fault location accuracy. Traveling waves propagate at near the speed of light after originating at the fault point. Therefore, the arrival time of the traveling wave at both ends must be accurately detected within an extremely short time. Thus, a sampling rate of 1MHz is used at both detection points, with a sampling interval of 1s, to ensure more accurate traveling wave signal detection. To more accurately calculate the time difference between the initial traveling wave arrival at detection points A and B, the Teager-Kaiser energy operator is used to process the line-mode traveling wave signal. A threshold method is used to set a value suitable for DC transmission line faults. Under the condition of meeting this threshold, the sampling points of the first energy jump at both ends of the traveling wave signal are extracted. Then, combined with the sampling rate at both ends of the line, the time difference between the fault traveling wave signal and the arrival time at detection points A and B is accurately calculated. .

[0015] Figure 6 This is a control flowchart for fault identification and location. A fault location method based on wavelet decomposition and TKEO is proposed, which consists of three stages: 1) fault data processing; 2) fault identification; and 3) fault location.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0017] 1. In terms of fault identification, it can correctly identify three situations: positive grounding fault, lightning interference, and lightning fault. The noise reduction effect is excellent, and it can accurately identify the fault at signal-to-noise ratios of 20dB, 30dB, and 40dB.

[0018] 2. In terms of fault location, it can still accurately locate faults at different distances, ensuring that the error is within 200m, and the ranging accuracy is not affected under high transition resistance.

[0019] 3. The ranging results of this method are largely unaffected by different transition resistances and fault distances, fully verifying its high accuracy. Furthermore, the ranging accuracy remains high for different fault types, meaning that this method can reduce reliance on hardware devices in practical applications, improving the system's economy and practicality. Attached Figure Description

[0020] Figure 1 This is a waveform diagram of lightning current in this invention;

[0021] Figure 2 This is a comparison chart of the spectral analysis of different decomposition layers in this invention;

[0022] Figure 3 is a schematic diagram of the zero-mode component waveform of the positive grounding fault in this invention, wherein Figure 3(a) is the current waveform of the positive grounding fault and Figure 3(b) is the spectrum characteristic of the positive grounding fault.

[0023] Figure 4 is a schematic diagram of the zero-mode component waveform of lightning interference in this invention, wherein Figure 4(a) is a waveform diagram of lightning interference current and Figure 4(b) is a waveform diagram of lightning interference spectrum characteristics.

[0024] Figure 5 is a schematic diagram of the zero-mode component waveform of the lightning fault in this invention, wherein Figure 5(a) is the lightning fault current waveform and Figure 5(b) is the lightning fault spectrum characteristic.

[0025] Figure 6 This is a control flowchart for fault identification and location in this invention;

[0026] Figure 7 This is a lightning current model diagram from an embodiment;

[0027] Figure 8 These are the simulation model parameters in the embodiments;

[0028] Figure 9 This is a graph showing the K-values ​​under different fault types in the embodiments;

[0029] Figure 10 These are the energy values ​​of each decomposition layer in the embodiment;

[0030] Figure 11 These are the ranging results under different transition resistances in the embodiments;

[0031] Figure 12 This is a comparison chart of errors caused by lightning strikes at different distances in the embodiments;

[0032] Figure 13 This is a comparison chart of ranging time with other methods in the embodiment. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] In the fields of lightning protection engineering and high-voltage insulation testing, the 1.2 / 50s lightning impulse current waveform is widely adopted as an internationally standardized test waveform to simulate the waveform characteristics of actual lightning impulse currents, ensuring that lightning protection devices can effectively cope with lightning strikes. The main discharge phase of a lightning strike can be figuratively described as the process by which a lightning wave originates from a thundercloud and propagates rapidly along a theoretically infinitely extending lightning channel with a fixed wave impedance Z towards a predetermined strike point. In this invention, the following method is used... Figure 7 The lightning current model is given, where the lightning channel impedance is 300Ω.

[0035] When decomposing the signal, the db4 wavelet is selected as the wavelet basis and the number of decomposition layers is 11. Wavelet decomposition is performed on the zero-mode component of the fault signal, and the energy of different decomposition layers is calculated. The fault type is determined according to the magnitude of the energy ratio K. Wavelet decomposition is performed on the linear-mode component of the fault signal, and then TKEO is used to extract the wavefront and accurately calculate the fault distance. Figure 8 These are the parameters for the simulation model.

[0036] Figure 9 This is a graph showing the ratio of high-frequency layer energy to low-frequency layer energy for lightning strikes, positive grounding faults, and lightning interference at fault distances of 20, 30, 40, and 50 km.

[0037] Taking the occurrence of lightning strike, lightning interference, and positive grounding fault at a distance of 45km as examples, wavelet decomposition is performed on the zero-mode current. Based on the corresponding formulas, the energy values ​​of each layer, the high-frequency layer, the low-frequency layer, and the K value are obtained. The energy value of the high-frequency layer is the sum of the energies of layers 1-8, and the energy value of the low-frequency layer is the sum of the energies of layers 9-11. The energy ratio K is the ratio of the energy of the high-frequency layer to that of the low-frequency layer. The energy of each decomposed layer is as follows: Figure 10As shown in the figure. Due to the influence of the transition resistance, the propagation of the fault traveling wave undergoes a faster attenuation process. This phenomenon makes the characteristics of the fault traveling wave blurred and difficult to identify, increasing the difficulty of fault location. To verify the influence of the transition resistance on the ranging accuracy of the proposed method, the transition resistance was set to 0.1, 100, and 300Ω, respectively, and simulation verification was carried out at fault locations of 26, 51, and 71km for fault types of lightning strike fault and positive grounding fault. The results are shown in the figure. Figure 11 As shown.

[0038] To verify the superior ranging accuracy of the proposed method, simulations were conducted to compare it with ranging methods based on EMD-TEO and VMD-TEO. The VMD algorithm was configured with 5 decomposition layers and a penalty factor of 2500. The EMD algorithm adaptively determines the number of decomposition layers based on the characteristics of the signal itself, and to avoid over-decomposition and feature aliasing, a maximum decomposition layer of 12 was set.

[0039] The error ratio of lightning strike failures occurring at different distances is as follows: Figure 12 As shown, simulation results indicate that the EMD-TEO ranging method has higher ranging accuracy when the fault point is closer to the monitoring points at both ends, while the VMD-TEO ranging method has higher ranging accuracy when the fault point is closer to the middle of the line. This method has advantages in both the ends and the middle, so this method has higher accuracy in fault location.

[0040] Under the same experimental conditions, the time consumption of this method was compared with that of the EMD-TEO algorithm and the VMD-TEO algorithm. The specific results are as follows: Figure 13 As shown, this method has a significant advantage over the EMD-TEO and VMD-TEO algorithms in terms of ranging time. This indicates that, under the same conditions, this method can complete the ranging task more efficiently and in less time, further validating its potential and advantages in practical applications.

Claims

1. A fault identification and location method based on wavelet transform and TKEO, the method being applied to an 800V photovoltaic DC hydrogen production system, characterized in that: The accuracy of the proposed method was verified and analyzed for three scenarios: positive grounding fault, lightning interference, and lightning strike fault. The fault identification method incorporates the high-low frequency decomposition layer ratio to differentiate between these three scenarios. For fault location, the TKEO method is used to calibrate the wavefront, and then the fault distance is calculated using the two-end traveling wave ranging method. Considering that transmission lines are laid in natural environments and are susceptible to noise, soft thresholding is used for denoising, followed by wavelet transform for signal reconstruction and further denoising. Noise signals contain a large number of high-frequency signals. The soft thresholding method is used to process the wavelet decomposition result, and the formula is as follows: In the formula: These are the denoised wavelet coefficients; The wavelet coefficients before denoising are: T = 1; T is the threshold value. Represents a symbolic function. The threshold T is a general threshold, and the calculation formula is: In the formula: l is the length of the fault signal; The standard deviation of the noise can be estimated using the median of the wavelet high-frequency coefficients. In the formula: W represents the high-frequency wavelet coefficients.

2. In the fault identification method as described in claim 1, the energy ratio K of different decomposition layers is introduced as a criterion for fault identification, where D1-D8 are high-frequency layers and D9-D11 are low-frequency layers. The energy calculation formula for each decomposition layer is as follows: In the formula: The wavelet energy of the Nth layer; The wavelet coefficients of the Nth layer after wavelet decomposition of the zero-mode component of the fault current. and These represent the energies of the high-frequency and low-frequency layers after wavelet decomposition, respectively. Combining a certain margin, we can conclude that: when When the fault type is determined to be a positive ground fault, then the fault type is determined to be a positive ground fault. When, it is determined to be lightning interference; when At that time, the fault type was determined to be a lightning strike fault.

3. According to the fault location method in claim 1, using the TKEO calibrated wavefront, the Teager-Kaiser energy operator expression for a continuous-time signal x(t) is: In the formula: and These are the first and second derivatives, respectively. For a discrete-time signal x(n), the Teager-Kaiser energy operator expression is: In the formula: These are discrete time signal sampling points; This represents the energy value at the sampling point. Using the time difference of signal detection at both ends for calculation can better address the attenuation and distortion problems that may occur during the propagation of traveling wave signals. The principle formula of the two-end traveling wave ranging method is as follows: In the formula: and These are the sampling points at the left and right ends of the detection point where the initial traveling wavefront arrives; The sampling rate at both detection points; , These represent the distances from the fault point to both ends of the detection point; L is the distance between the detection points at both ends of the line; and v is the traveling wave velocity of the line model. This is the time difference between the arrival of the initial traveling wave front at both ends of the fault. In traveling wave ranging, the wave velocities of zero-mode traveling waves differ significantly among their frequency components, resulting in pronounced wave velocity variation characteristics. In contrast, the wave velocity of linear mode traveling waves is much less affected by frequency variations than that of zero-mode traveling waves. Since linear mode traveling waves experience less attenuation during propagation, their wave velocity is relatively stable. Therefore, in ranging research, linear mode traveling waves are analyzed, and their velocity is used to locate fault points. The formula for calculating the traveling wave velocity of a linear mode is: In the formula: , , , These represent the resistance, inductance, conductance, and capacitance per unit length of a line at a given frequency. This is the angular frequency of the frequency component. The conductivity of a typical power transmission line can be ignored, that is... Then the above formula can be simplified to: Given transmission line parameters, the wave velocity varies with frequency. When the frequency exceeds 100Hz, the wave velocity tends to be constant. However, the frequency of traveling waves during faults is often between several kHz and several hundred kHz. Therefore, the influence of frequency on wave velocity can be ignored when studying traveling wave fault location. That is, when the frequency... When it is very large, R0 / Approximately zero, the above formula can be further simplified to: At this point, the traveling wave velocity of the linear mode is constant, where , The traveling wave velocity of the line model can be obtained by substituting the parameters into the above formula. .

Images