Hybrid multi-terminal direct current transmission line fault location method based on wave velocity scaling factor
By combining robust local mean decomposition and multi-resolution singular value decomposition, and utilizing wave velocity scaling factor and time difference characteristics, the problem of wavefront calibration and wave velocity selection in hybrid multi-terminal DC transmission systems is solved, achieving fast and accurate fault location, and is suitable for complex topologies and high-noise environments.
Patent Information
- Application Number
- CN202411054555.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
AI Technical Summary
Existing fault location methods for DC transmission systems suffer from difficulties in wavefront calibration, wave velocity selection, and large ranging errors due to complex topology in hybrid multi-terminal DC transmission systems. In particular, they are difficult to achieve fast and accurate fault location in transmission lines with complex terrain and high fault probability.
A method based on Robust Local Mean Decomposition (RLMD) and Multi-Resolution Singular Value Decomposition (MRSVD) is adopted, combined with the wave velocity scaling factor, and the traveling wave signal is decoupled by Clark transform to construct the Hankel matrix, extract the detail components, fit the curve using PCHIP interpolation, identify the fault branch by combining the time difference characteristics, and solve the fault distance simultaneously.
It achieves fast and accurate fault location in hybrid multi-terminal DC transmission systems, reduces computational load, improves location accuracy, reduces the impact of synchronization errors and noise interference, is suitable for complex topologies and high-noise environments, and has strong resistance to transition resistance.
Smart Images

Figure CN118980885A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of power grid fault detection, in particular to a hybrid multi-terminal direct current transmission line fault location method based on a wave velocity proportional factor. Background Art
[0002] In the context of the reverse distribution of energy supply and load demand in my country, high-voltage direct current transmission is an important technical support for promoting the rational allocation of resources, optimizing the energy structure, and promoting the clean and low-carbon development of electricity. Among the existing direct current transmission methods, the direct current transmission technology based on line commutated converter (LCC) is mature, but the inverter side is prone to commutation failure; while the direct current transmission technology based on modular multilevel converter (MMC) has no commutation failure problem, but it is difficult to increase the capacity; for this reason, hybrid multi-terminal direct current transmission combining the advantages of LCC and MMC has become a new means of long-distance and large-capacity power transmission in recent years due to its flexible operation mode, low transmission loss, and large capacity, and has been successfully applied in the Kunliulong UHV hybrid multi-terminal direct current transmission project. However, due to its long transmission distance, complex terrain, high probability of direct current line failure, and difficulty in manual line inspection, it still faces technical difficulties in the field of fault location. Therefore, it is necessary to study a fast and accurate fault location method to guide fault troubleshooting and ensure that the power system quickly recovers to a stable state.
[0003] The existing DC transmission fault location methods mainly include three categories: traveling wave method, natural frequency method and fault analysis method. Among them, the traveling wave method has been widely studied and applied because it is not easily affected by the operation mode and does not require the extraction of natural frequency. The key lies in the calibration of the traveling wave head and the selection of wave velocity.
[0004] In order to solve the problem of wave head calibration, the existing wave head calibration methods are mostly based on algorithms such as wavelet transform, empirical mode decomposition, and Hilbert transform. The above methods are generally complex to calculate, and the Hilbert transform has the problem of negative frequency. The improved empirical mode algorithm lacks a clear theoretical basis and the algorithm parameters are difficult to select, which has a great impact on the calibration results. In order to solve the problem of difficulty in selecting wave velocity, the existing two-terminal ranging methods mostly eliminate the wave velocity variable in the ranging formula to avoid the selection of wave velocity. It ignores the speed difference of the traveling wave propagating to the two ends of the line. When the fault is close to the line boundary, the above methods may have large errors. In addition, for hybrid multi-terminal DC transmission systems with a large number of branches and complex topological structures, it is difficult to calibrate the refractive wave head due to the influence of adjacent lines. The existing ranging scheme that eliminates the influence of wave velocity by calibrating the reflected wave is difficult to directly apply in hybrid multi-terminal systems. Summary of the invention
[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor. By combining robust local mean decomposition and multi-resolution singular values, wave head calibration with zero offset advantage is achieved. In addition, on the basis of using the time difference characteristics of a single measuring point to identify the fault branch, fast and accurate fault location is achieved by solving equations simultaneously.
[0006] The present invention solves the technical problem by adopting the following technical solutions:
[0007] A hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor comprises the following steps:
[0008] Step 1: simulate and sample multiple groups of fault voltage traveling wave data samples of known fault points, and use Clark transformation to decouple them into line mode and zero mode components;
[0009] Step 2: Extract the PF1(t) component of the fault voltage traveling wave based on the robust local mean decomposition (RLMD) and construct the Hankel matrix;
[0010] Step 3: Extract the second-layer detail component D2 of PF1(t) based on multi-resolution singular value decomposition (MRSVD) to calibrate the time when the line mode and zero mode traveling waves arrive at the two end measuring points;
[0011] Step 4: Calculate the α corresponding to different fault points based on the sample calibration results, and use PCHIP interpolation to solve the α(x) fitting curve;
[0012] Step 5, sampling the fault data of the fault point to be found and calibrating its traveling wave arrival time;
[0013] Step 6: Determine the fault line according to the fault branch judgment criterion;
[0014] Step 7: Based on the calibration result of the fault point to be determined, the α(x) calculation curve is obtained, and the fault distance is obtained by solving it together with the fitting curve.
[0015] Moreover, the specific implementation method of step 2 is: the robust local mean decomposition RLMD adaptively decomposes the fault voltage traveling wave X(t) into a series of product function components PF and a residual component ε(t):
[0016]
[0017] Among them, K is the number of decompositions, PF k (t) represents the kth PF component;
[0018] Construct the Hankel matrix H using components PF1 = [PF1(1),…,PF1(N)] of length N:
[0019]
[0020] Moreover, the specific implementation method of step 3 is: multi-resolution singular value decomposition MRSVD separates the Hankel matrix H into approximate components and detail components:
[0021] H=PSQ T =H a +H d
[0022] Where P and Q are left and right orthogonal matrices respectively; S = diag(σ a ,σ d ) is a singular value matrix, where the large σ a The singular values of the corresponding approximate components, small values of σ d The singular value corresponding to the detail component; H a , H d ∈B 2×(N-1) are the approximate matrix and the detail matrix respectively, and the approximate matrix H a and the detail matrix H d After one layer of decomposition, we get the approximate component A1 and the detail component D1, where A1 reflects the overall trend of the signal and D1 reflects the detail information of the signal, which is used to extract the traveling wave head. A1 is used to construct the Hankel matrix again, and the above steps are repeated S times to obtain a multi-scale decomposition signal including S approximate components and S detail components. By directly observing the waveform of the detail component, the time corresponding to the modulus maximum point is determined, which is the arrival time of the desired traveling wave.
[0023] Moreover, the specific implementation method of step 4 is: for the multiple sample sets with known fault distances, the wave velocities of different components are calculated in combination with the calibrated traveling wave arrival times, and the corresponding α is solved:
[0024] α=(1 / v M1 -1 / v M0 ) / (1 / v N1 -1 / v N0 )
[0025] For the α corresponding to different fault locations calculated by the above formula, the α(x) fitting curve is solved based on the piecewise cubic Hermit interpolation PCHIP: According to the given n interpolation nodes e i (i=1,...,n) and the corresponding function value y i , based on MATLAB, the first-order derivative m that correctly reflects the shape and change trend of discrete points is estimated i , which is then used to construct a piecewise cubic Hermite interpolation function H(e) to satisfy H(e i )=yi , H′(e i )=m i , and H(e) is in every interval [e i ,e i+1 ] are all cubic polynomials, and their interpolation basis functions are:
[0026]
[0027] in,
[0028] Furthermore, the implementation method of step 5 is: when a fault occurs in real time on the line, the fault voltage traveling wave corresponding to the fault point to be found is collected, and the traveling wave time is calibrated using the same calibration method as step 2-3.
[0029] Moreover, the fault branch criterion in step 6 is:
[0030]
[0031] Δt MN =t MN1 -t MN0
[0032] Δt M =Δt MN -Δt N
[0033] Among them, Δt MN is the time difference between the line mode and zero mode at both ends when propagating on line l1, t MN1 ,t MN0 They represent the time required for the line mode and zero mode traveling waves to propagate along the entire length of line l1, Δt M is the time difference between the line mode and zero mode traveling waves arriving at the M measuring point, Δt N is the time difference between the line mode and zero mode traveling waves arriving at N measuring points.
[0034] Moreover, the specific implementation method of step 7 is: solving the α(x) calculation curve based on the traveling wave calibration result of step 5:
[0035]
[0036] Wherein, L1 represents the length of line l1.
[0037] The fault distance can be obtained by solving the equation by combining the calculated curve with the fitting curve obtained in step 4.
[0038] The advantages and positive effects of the present invention are:
[0039] 1. The present invention analyzes the variation characteristics of wave velocity along the line, makes full use of the characteristics that the wave velocities of different moduli at the same measuring point have the same variation trend but different variation speeds, and defines a wave velocity proportional factor that can indirectly reflect the fault distance. Secondly, the fitting curve and calculation curve of the proportional factor changing with the fault distance are solved by the interpolation fitting and quantitative derivation methods respectively, and the distance measurement scheme of the fault distance is proposed by using the monotonicity difference. In order to accurately calibrate the fault traveling wave, a wave head calibration method combining robust local mean decomposition and multi-resolution singular value decomposition is proposed, and the calibration results are used to design a fault branch identification criterion based on the time difference characteristics. This method has high ranging accuracy, small calculation amount, is not affected by synchronization errors, and has strong resistance to transition resistance (500Ω) and noise (20dB), and has strong applicability in different lines.
[0040] 2. The present invention introduces traveling wave velocities of different directions and moduli by defining a wave velocity proportional factor, thereby eliminating the influence of synchronization errors and overcoming the defect of large ranging errors caused by the traditional traveling wave ranging method due to ignoring the difference in wave velocity propagating to both ends of the line.
[0041] 3. The traveling wave calibration method based on RLMD-MRSVD of the present invention has the advantages of small calculation amount and zero offset of singular points. The fault branch identification criterion designed based on the calibration result only uses the time difference characteristics of a single measuring point under different line faults, has low data volume requirements and is not easily affected by synchronization errors and operating modes.
[0042] 4. The present invention is based on the distance measurement method proposed in the present invention. After the fault occurs, it only needs to calibrate the fault traveling wave and solve a simple set of linear equations in parallel to achieve fault distance measurement. It has high accuracy, small calculation amount and fast calculation speed, and overcomes the shortcomings of the iterative distance measurement method with many calculation times and the high sample requirement of the distance measurement method based on artificial intelligence algorithm.
[0043] 5. The present invention does not require calibration of the catadioptric wave head and has strong applicability in hybrid multi-terminal DC transmission systems with complex topology. The simulation results show that the proposed scheme has high ranging accuracy under different fault conditions, is less affected by noise (20dB), has strong resistance to transition resistance (500Ω), and still has higher accuracy than the traditional two-terminal traveling wave ranging method when the line length changes and the line parameters fluctuate. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a structural diagram of the present invention.
[0045] Figure 2 It is a simplified model diagram of a hybrid three-terminal DC transmission system;
[0046] Figure 3 It is a schematic diagram of the line mode and zero mode traveling wave transmission path;
[0047] Figure 4 It is the α(x) fitting curve diagram corresponding to the line l1 of the present invention;
[0048] Figure 5 A schematic diagram of a simulation model constructed for an embodiment of the present invention;
[0049] Figure 6 Schematic diagram of calibration results of different methods in embodiments of the present invention;
[0050] Figure 7 This is a schematic diagram of comparing errors of different distance measurement methods according to an embodiment of the present invention;
[0051] Figure 8 is a fitting curve diagram of α2(x) corresponding to line l2 of an embodiment of the present invention;
[0052] Fig. 9 The figure is a comparison of errors of different methods under clock asynchrony in an embodiment of the present invention. DETAILED DESCRIPTION
[0053] The present invention is further described in detail below with reference to the accompanying drawings.
[0054] A hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor, such as Figure 1 As shown, steps 1 to 4 are the offline part, and steps 5-7 are the online part, which includes the following steps:
[0055] Step 1: simulate and sample multiple groups of fault voltage traveling wave data samples of known fault points, and use Clark transform to decouple them into line mode and zero mode components.
[0056] Step 2: Extract the PF1(t) component of the fault voltage traveling wave based on the robust local mean decomposition (RLMD) and construct the Hankel matrix.
[0057] Based on LMD, RLMD optimizes the shortcomings of LMD in terms of boundary conditions, envelope estimation and screening stop criteria. It can adaptively decompose the original signal X(t) into a series of product function components (PF) and a residual component ε(t), as shown in formula (1).
[0058]
[0059] Where K is the number of decompositions, PF k (t) represents the kth PF component.
[0060] For each PF component obtained by decomposition, its essence is a single-component amplitude-frequency modulated signal, which can be represented by the product of the envelope signal and the frequency-modulated signal. The instantaneous amplitude of PF corresponds to the envelope signal in the product term, and the instantaneous frequency can be solved by the frequency-modulated signal. Since it is independent of the instantaneous amplitude, there is no need to perform Hilbert transform, thus avoiding the negative frequency problem of EMD. In addition, RLMD solves the envelope estimation function based on the sliding average method during the decomposition process, overcoming the modal aliasing defect of EMD using cubic spline function to fit the envelope line. Compared with the intrinsic mode function obtained by EMD decomposition, the physical meaning of the product function PF is clear and can more intuitively reflect the inherent characteristics of the signal.
[0061] Step 3: Extract the second-layer detail component D2 of PF1(t) based on multi-resolution singular value decomposition (MRSVD) to calibrate the time when the line mode and zero mode traveling waves arrive at the two end measuring points.
[0062] MRSVD is a multi-resolution decomposition algorithm based on binary recursive decomposition, which has good applications in detecting signal mutation characteristics and noise reduction. When decomposing, it first uses the component PF1=[PF1(1),…,PF1(N)] of length N to construct the Hankel matrix H shown in formula (2):
[0063]
[0064] It can be seen from formula (2) that the two rows of the H matrix are highly correlated. By performing SVD on them, the approximate component and the detail component can be separated as shown in formula (3).
[0065] H=PSQ T =H a +H d (3)
[0066] Where P and Q are left and right orthogonal matrices respectively; S = diag(σ a ,σ d ) is a singular value matrix, where the larger σ a The singular values of the corresponding approximate components, the smaller σ d The singular value corresponding to the detail component; H a , H d ∈B 2×(N-1) They are the approximate matrix and the detail matrix respectively.
[0067] Based on the above H a , H dAfter one layer of decomposition, we can get the approximate component A1 and the detail component D1, where A1 reflects the overall trend of the signal and D1 reflects the detail information of the signal, which can be used to extract the traveling wave head. Similarly, we use A1 to construct the Hankel matrix again, repeat the above steps S times, and we can get a multi-scale decomposition signal including S approximate components and S detail components.
[0068] From the above analysis, it can be seen that MRSVD, similar to wavelet transform, can decompose the signal into a linear combination of approximate components and detail components. However, when using wavelet transform to detect singularity, the detected singular points will shift with the wavelet basis and decomposition scale. In contrast, MRSVD does not need to select wavelet basis functions, and the positions of singular points under different decomposition scales are the same, so each layer of detail components can be used to calibrate the traveling wave head.
[0069] Compared with the wave head detection method based on EMD and wavelet transform, the traveling wave calibration method combining the advantages of RLMD-MRSVD does not require Hibert transform and FFT transform, has a small amount of calculation, a faster operation speed, and is more conducive to the extraction of fault features. Accordingly, when calibrating the traveling wave head, the present invention first uses RLMD to extract the time-frequency characteristics of the sampled fault traveling wave signal, and selects the PF1 component to construct the Hankel matrix. Then, MRSVD is used to separate the detail components in PF1. Considering that the noise in the transmission circuit will interfere with the detail components, and as the MRSVD decomposition scale increases, the noise ratio in the detail components will gradually decrease, and the amount of calculation required for ranging will increase. Taking the above factors into consideration, the present invention takes S=2, that is, D2 is selected to calibrate the arrival time of the traveling wave, and the time corresponding to the modulus maximum point is determined by directly observing its waveform. This time is the desired arrival time of the traveling wave.
[0070] Step 4: Calculate the α corresponding to different fault points based on the sample calibration results, and use PCHIP interpolation to solve the α(x) fitting curve.
[0071] The wave velocity proportional factor α that reflects the wave velocity variation characteristics of different moduli and directions is shown in formula (4):
[0072] α=(1 / v M1 -1 / v M0 ) / (1 / v N1 -1 / v N0 ) (4)
[0073] Since the zero-mode wave velocity changes more significantly with fault distance than the line-mode component, it can be seen from formula (4) that the overall change trend of α is mainly determined by (1 / v M0 ) / (1 / v N0 ) determines that when x increases, v M0 Decrease, vN0 As α increases, α also increases. This means that on the coordinate plane with α as the ordinate and x as the abscissa, α(x) is a monotonically increasing curve, and there is a one-to-one correspondence between α and x. Therefore, the size of α can indirectly reflect the fault distance x.
[0074] For multiple sets of samples with known fault distances, the wave velocities of different components are calculated by combining the calibrated traveling wave arrival times, and the corresponding α is solved using the above formula. Furthermore, the α(x) fitting curve can be solved based on the piecewise cubic Hermit interpolation PCHIP: i (i=1,...,n) and the corresponding function value y i , based on MATLAB, the first-order derivative m that correctly reflects the shape and change trend of discrete points is estimated i , which is then used to construct a piecewise cubic Hermite interpolation function H(e) to satisfy H(e i )=y i , H′(e i )=m i , and H(e) is in every interval [e i ,e i+1 ] are all cubic polynomials, and their interpolation basis functions are:
[0075]
[0076] in,
[0077] Step 5: Sample the fault data of the fault point to be found and calibrate its traveling wave arrival time.
[0078] When a fault point to be found occurs in real time on the line, the corresponding fault voltage traveling wave is collected, and the traveling wave time is calibrated using the calibration method of step 2-3.
[0079] Step 6: Determine the faulty line according to the fault branch judgment criteria.
[0080] like Figure 2 This is a simplified topology model of a hybrid three-terminal DC transmission system. Compared with conventional two-terminal DC transmission, its structure is more complex and the line contains branches. Therefore, before fault location measurement, the fault branch needs to be identified first.
[0081] Define the time difference Δt between the line mode and zero mode at both ends when the traveling wave propagates on branch l1 MN As shown in formula (6):
[0082] Δt MN =t MN1 -t MN0 (6)
[0083] Where, t MN1 ,t MN0 They respectively represent the time required for line mode and zero mode traveling waves to propagate along the entire length of line l1.
[0084] If the fault point is located on branch l1, then equation (7) holds:
[0085]
[0086] Combining equations (6) and (7), we can get the time difference Δt between the line mode and zero mode traveling waves arriving at the M measuring point: M As shown in formula (8):
[0087] Δt M =Δt MN -Δt N (8)
[0088] Since the zero-mode wave velocity is less than the line-mode wave velocity, that is, Δt N <0, then Δt M >Δt MN .
[0089] On the contrary, if the fault point is located on branch l2, equation (9) holds:
[0090]
[0091] Combining equation (6) and equation (9), Δt N It can be expressed as shown in formula (10):
[0092] Δt M =Δt MN +Δt N (10)
[0093] At this time, Δt M <Δt MN .
[0094] In summary, the fault branch criterion can be expressed as shown in formula (11):
[0095]
[0096] In the formula, Δt MN It is the setting value of the branch criterion, which can be calculated from the arrival time of the line and zero mode traveling waves measured at M when the fault occurs in the T zone. It can be seen from formula (11) that the fault branch criterion proposed in the present invention only needs to obtain the fault information of two measuring points. Since each measuring point introduces the line mode and zero mode characteristic quantities, the criterion is also not affected by the synchronization error.
[0097] Step 7: Obtain the α(x) calculation curve and solve it together with the fitting curve to obtain the fault distance.
[0098] In order to analyze the quantitative relationship between α and the calculation results of fault distance under a certain fault position, taking the fault of line l1 at time t0 as an example, the line mode and zero mode fault traveling wave transmission paths are drawn as follows: Figure 3 As shown in the figure, t M0 ,t N0 are the moments when the zero-mode traveling wave at the fault point reaches the M and N measuring points, respectively, satisfying equation (12):
[0099]
[0100] Where L1 is the length of line l1.
[0101] Then define the line and zero mode time difference Δt at M and N M , Δt N As shown in formula (13):
[0102]
[0103] Substituting formula (13) into formula (12), we can get:
[0104]
[0105] Combining equation (12) and equation (14), the fault distance calculation expression can be obtained as shown in equation (15):
[0106]
[0107] According to the above formula, at a certain fault location, Δt M , Δt N As a constant, the calculated result of the fault distance x will gradually decrease with the increase of α, and a monotonically decreasing α(x) curve can be determined accordingly. On this basis, the analysis of formula (4) has shown that there is a unique α corresponding to the fault position. Therefore, the process of solving the fault distance at this time is essentially to determine a set of α and x on the increasing α(x) curve so that it can simultaneously satisfy the α(x) decreasing curve determined by formula (15), that is, the x value corresponding to the intersection of the two curves is the actual fault distance.
[0108] In summary, in order to achieve fault distance measurement, the calculation curve and fitting curve of α(x) can be used to solve their intersection. The fitting curve is obtained by solving steps 1-4, which is applicable to the fault point at any position on the same line. Therefore, the above steps only need to be executed once for one line, and can be executed offline before the fault occurs. The calculation curve is obtained by steps 5-7, which needs to be solved based on the fault information of the fault point to be found in real time, and needs to be executed online after the fault occurs. It can be seen that whenever a fault occurs on the line, only the online part of the steps needs to be executed to achieve fault distance measurement, which reduces the amount of real-time calculation and realizes fast distance measurement.
[0109] Based on the above method, for the simulation model adopted by the present invention, taking line l1 as an example, a single-pole ground fault is set as an interpolation node every 100 km, and the corresponding α value is calculated by combining equations (4) and (15) and fitted using PCHIP. The α(x) characteristic curve obtained by solving is as follows: Figure 4 .
[0110] Depend on Figure 4 It can be seen that α gradually increases with the increase of fault distance. The physical meaning of the α(x) fitting curve with increasing characteristics is essentially to describe the corresponding relationship between each wave velocity and the fault distance.
[0111] According to the hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor, calculations are performed by referring to the model built for the Kunliulong hybrid three-terminal DC transmission project to verify the effect of the present invention.
[0112] Referring to the Kunliulong hybrid three-terminal DC transmission project, the following is constructed based on the PSCAD / EMTDC platform: Figure 5 The simulation model shown: the system rated voltage is ±800kV; the DC transmission line adopts a frequency-dependent phase domain model, where line l1 is 942km long and has a rated current of 5kA; line l2 is 532km long and has a rated current of 3kA.
[0113] For long-distance high-voltage direct current transmission systems, the sampling frequency of the traveling wave ranging device is usually not less than 1 MHz. The present invention sets the sampling frequency to 500 kHz to verify the reliability of the proposed ranging scheme when the sampling frequency is low.
[0114] In order to verify the feasibility of the traveling wave calibration method based on RLMD-MRSVD, the fault occurrence time t0=0.8s, the duration 0.5s, and the fault distance x=500km are set. Different wave head detection methods are used to calibrate the arrival time of the fault traveling wave to compare their effects. The simulation results are shown in Figure 2. Figure 6 shown.
[0115] Depend on Figure 6 (a) It can be seen that as the decomposition scale increases, the detail components D1-D3 extracted by the traveling wave calibration algorithm proposed in the present invention accurately intersect at the singular point, and the pulses are approximately centrally symmetrically distributed with the intersection point, and the pulse width increase is small. Figure 6 (b) and Figure 6 (c) It can be seen that the components d1-d3 obtained by decomposition of db wavelet and sym wavelet do not intersect at the singular point due to offset, and their pulse width increases significantly with the increase of decomposition scale, which may cause the reflected wave to be submerged by the pulse width of the initial wave head and cannot be calibrated, making the existing ranging scheme that uses reflected waves to eliminate the influence of wave velocity invalid.
[0116] In summary, the traveling wave calibration method based on RLMD-SVD overcomes the defect that the singular points of wavelet transform shift with scale, and can realize accurate calibration of traveling waves at different scales.
[0117] In order to verify the reliability of the fault branch judgment criterion proposed in the present invention, different lines are set to have faults, and the line zero mode time difference Δt when the traveling wave reaches M is calculated. M , the simulation results are shown in Table 1.
[0118] Table 1 Calculation results of fault branch judgment criteria
[0119]
[0120]
[0121] After simulation test, when the fault occurs in the T zone, the Δt measured at M is M =0.34ms, Δt N =0ms, i.e. fault branch criterion setting value Δt MN =Δt M =-0.34ms. From Table 1, we can see that when the fault is located in branch l1, Δt M Both are greater than the set value Δt MN ; When the fault is located in branch l2, Δt M Both are less than the set value Δt MN , which satisfies the identification criterion shown in formula (11), indicating that the proposed criterion is capable of identifying the faulty branch.
[0122] In order to fully verify the reliability and accuracy of the proposed ranging scheme, taking line l1 as an example, the ranging results under different fault conditions are simulated and analyzed, as shown in Table 2. It is also compared with the traditional two-terminal ranging method and the existing improved two-terminal traveling wave ranging method based on catadioptric wave. The relative error is shown in Table 2. Figure 7 .
[0123] Table 2 Ranging results under different fault conditions
[0124]
[0125] Depend on Figure 7It can be seen that the traditional two-terminal ranging method uses the same wave velocity for the entire line, so the relative error of the calculation result is large and the ranging accuracy is low. The existing improved two-terminal ranging scheme that does not require wave velocity selection ignores the difference in wave velocity of the traveling wave propagating to the two ends of the line. It can be seen that when the fault is close to the two ends of the line, the speed difference of the traveling wave propagating in different directions is large, and the ranging error of this method increases. In addition, since it needs to calibrate multiple reflected waves, it has poor adaptability in a multi-branch hybrid multi-terminal system. According to the results in Table 2, the ranging error of the ranging scheme proposed in the present invention is less than 0.4% at different fault distances, and changes little with the increase of transition resistance. When the transition resistance is 500Ω, the ranging error is less than 1%. Compared with the traditional two-terminal traveling wave ranging and the improved two-terminal ranging method, the error is reduced, and it does not need to calibrate the reflected wave and is not affected by the complex topology of the system.
[0126] In addition to the influence of line length, in actual engineering, due to the influence of seasons, environment and other factors, the parameters of the line will fluctuate, causing the α(x) fitting curve to shift compared with the typical operation, thus affecting the distance measurement results. Therefore, simulation tests need to be performed under different line parameters and length conditions. Since the transmission line l2 in the research object of the present invention has a small rated current and a thin wire diameter, its corresponding wave impedance is calculated to increase by about 11% to 18% compared with line l1. It can be seen that there are differences in the lengths and parameters of the two lines. Therefore, taking line l2 as an example, the corresponding α2(x) curve is solved. The results are as follows: Figure 8 shown.
[0127] Depend on Figure 8 It can be seen that since the length of line l2 is smaller than that of l1, at the same fault distance, α2(x) corresponding to l2 is greater than α1(x) corresponding to l1. To further verify the reliability of the ranging based on the above fitting curve, the ranging results of line l2 at different fault distances are tested, as shown in Table 3.
[0128] Table 3 Ranging results of line l2
[0129]
[0130] It can be seen from Table 3 that when the line parameters and length change, the fault distance calculated by solving the fitting curve α2(x) corresponding to the line still has a small relative error and is more accurate than Figure 8 The traditional two-terminal traveling wave method shown in the figure shows that the proposed method is highly applicable to lines with different parameters and lengths.
[0131] At present, based on the GPS timing system, the synchronization error at both ends of the line is generally between 0 and 2μs. In order to analyze the impact of sampling clock asynchronization on the proposed ranging scheme, a synchronization error of ±2μs is set and the ranging results are calculated, as shown in Table 4. Fig. 9 shown.
[0132] Table 4 Ranging results under different synchronization errors
[0133]
[0134] From Table 4 and Fig. 9 It can be seen that when the synchronization error is ±2μs, the ranging error of the improved two-terminal traveling wave ranging method increases and is greatly affected by the synchronization error. However, the ranging results of the ranging scheme proposed in the present invention do not change under different synchronization errors. This is because the synchronization delay terms cancel each other out when calculating the time difference, Δt M , Δt N The size remains unchanged, which further makes the calculation curve of α(x) determined by substituting it into formula (15) unchanged, so the fault distance calculation result is not affected by the synchronization error.
[0135] In order to verify the anti-noise interference ability of the proposed ranging scheme, Gaussian white noise with different signal-to-noise ratios is added to the collected fault traveling waves. The ranging results are shown in Table 5.
[0136] Table 5 Ranging results under different signal-to-noise ratios
[0137]
[0138] It can be seen from Table 5 that with the increase of noise, the ranging error increases slightly compared with the results in Table 2, but the relative error is still within the range of 0.55%, indicating that the proposed ranging scheme has strong anti-noise ability.
[0139] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manner. Any other implementation manners derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.
Claims
1. A hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor, characterized in that: The following steps are involved: Step 1: simulate and sample multiple groups of fault voltage traveling wave data samples of known fault points, and use Clark transformation to decouple them into line mode and zero mode components; Step 2: Extract the PF1(t) component of the fault voltage traveling wave based on the robust local mean decomposition (RLMD) and construct the Hankel matrix; Step 3: Extract the second-layer detail component D2 of PF1(t) based on multi-resolution singular value decomposition (MRSVD) to calibrate the time when the line mode and zero mode traveling waves arrive at the two end measuring points; Step 4: Calculate the α corresponding to different fault points based on the sample calibration results, and use PCHIP interpolation to solve the α(x) fitting curve; Step 5, sampling the fault data of the fault point to be found and calibrating its traveling wave arrival time; Step 6: Determine the fault line according to the fault branch judgment criterion; Step 7: Based on the calibration result of the fault point to be determined, the α(x) calculation curve is obtained, and the fault distance is obtained by combining it with the fitting curve.
2. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 1 is characterized in that: The specific implementation method of step 2 is: the robust local mean decomposition RLMD adaptively decomposes the fault voltage traveling wave X(t) into a series of product function components PF and a residual component ε(t): Among them, K is the number of decompositions, PF k (t) represents the kth PF component; Construct the Hankel matrix H using components PF1 = [PF1(1),…,PF1(N)] of length N:
3. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 2 is characterized in that: The specific implementation method of step 3 is: multi-resolution singular value decomposition MRSVD separates the Hankel matrix H into approximate components and detail components: H=PSQ T =H a +H d Where P and Q are left and right orthogonal matrices respectively; S = diag(σ a ,σ d ) is a singular value matrix, where the large σ a The singular values of the corresponding approximate components, small values of σ d The singular value corresponding to the detail component; H a , H d ∈B 2×(N-1) are the approximate matrix and the detail matrix respectively, and the approximate matrix H a and the detail matrix H d After one layer of decomposition, we get the approximate component A1 and the detail component D1, where A1 reflects the overall trend of the signal and D1 reflects the detail information of the signal, which is used to extract the traveling wave head. A1 is used to construct the Hankel matrix again, and the above steps are repeated S times to obtain a multi-scale decomposition signal including S approximate components and S detail components. By directly observing the waveform of the detail component, the time corresponding to the modulus maximum point is determined, which is the arrival time of the desired traveling wave.
4. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 3 is characterized in that: The implementation method of step 4 is: for multiple sample sets with known fault distances, the wave velocities of different components are calculated in combination with the calibrated traveling wave arrival times, and the corresponding α is solved: α=(1 / v M1 -1 / v M0 ) / (1 / v N1 -1 / v N0 ) For multiple groups of α corresponding to different fault locations calculated by the above formula, the α(x) fitting curve is solved based on the piecewise cubic Hermit interpolation PCHIP: According to the given n interpolation nodes e i (i=1,...,n) and the corresponding function value y i , based on MATLAB, the first-order derivative m that correctly reflects the shape and change trend of discrete points is estimated i , which is then used to construct a piecewise cubic Hermite interpolation function H(e) to satisfy H(e i )=y i , H′(e i )=m i , and H(e) is in every interval [e i ,e i+1 ] are all cubic polynomials, and their interpolation basis functions are: in, 5. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 4 is characterized in that: The implementation method of step 5 is: when a fault occurs in real time on the line, the fault voltage traveling wave corresponding to the fault point to be found is collected, and the traveling wave time is calibrated using the same calibration method as step 2-3.
6. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 5 is characterized in that: The fault branch criterion in step 6 is: Δt MN =t MN1 -t MN0 Δt M =Δt MN -Δt N Among them, Δt MN is the time difference between the line mode and zero mode at both ends when propagating on line l1, t MN1 ,t MN0 They represent the time required for the line mode and zero mode traveling waves to propagate along the entire length of line l1, Δt M is the time difference between the line mode and zero mode traveling waves arriving at the M measuring point, Δt N is the time difference between the line mode and zero mode traveling waves arriving at the N measuring points.
7. The hybrid multi-terminal DC transmission line fault location method based on wave velocity proportional factor according to claim 6 is characterized by: The specific implementation method of step 7 is: solving the α(x) calculation curve based on the traveling wave calibration result of the fault point to be found: Wherein, L1 represents the length of line l1. The fault distance can be obtained by solving the equation by combining the calculated curve with the fitting curve obtained in step 4.
