A Fault Location Method for 35kV Power Grid Based on an Improved Prony Method

Abstract

This invention discloses a 35kV power grid fault location method based on an improved Prony method, comprising the following steps: After a power grid line fault occurs, the protection device disconnects the circuit breakers at both ends to isolate the faulty line; transient excitation signals generated by capacitor discharge are injected into both ends of the line through the fault location module, and multiple data acquisition units simultaneously record the damped oscillating current signal to obtain two sets of multi-measurement point time series data; the improved Prony algorithm combined with symplectic geometric mode decomposition is used to process the data, and after phase space reconstruction and signal denoising, characteristic parameters such as attenuation coefficient and angular frequency are identified; the characteristic parameters are substituted into the equations of the line fault physical model to solve the problem, and the location result is output after statistical correction of the fault distance results at multiple measurement points. This invention equates the fault network to a second-order RLC damped oscillating circuit, and completely eliminates the influence of unknown transition resistance by using the difference method of the two-end measurement data, thereby improving the location accuracy.

Description

Technical Field

[0001] This invention relates to the field of power system fault diagnosis technology, specifically a 35kV power grid fault location method based on the improved Prony method. Background Technology

[0002] As the core link connecting the main power grid and end users, the stability of the 35kV distribution network and the reliability of its power supply are of paramount importance. However, due to various factors such as line aging, external damage, and severe weather, faults in the 35kV distribution network occur frequently. These faults can easily lead to large-scale power outages for users and may even cause serious safety accidents such as equipment damage, electric shock, and fires. Therefore, after a fault occurs, the ability to efficiently and reliably detect the fault and quickly and accurately locate the fault point is of extraordinary significance for the power sector to quickly organize emergency repairs, eliminate safety hazards, and restore power supply to the affected area in a timely manner.

[0003] Currently, various fault location methods exist in the field of distribution network fault location, such as the traveling wave method, impedance method, and signal injection method. The traveling wave method calculates distance by capturing the time difference between the arrival of the high-frequency traveling wave signal generated at the moment of the fault at different monitoring points. This method is fast, but it has extremely high requirements for data synchronization and sampling rate, and is easily affected by system topology and noise interference. Especially in complex distribution networks, multiple reflections and refractions of the traveling wave can severely affect the accuracy of wavefront identification. The traditional impedance method estimates distance by measuring the voltage and current of the line to calculate the impedance, but its accuracy is severely affected by transition resistance, load current fluctuations, and uncertainties in line parameters.

[0004] Existing methods typically require complex analysis of massive amounts of transient data collected after a fault occurs, such as Short-Time Fourier Transform (STFT) and Wavelet Transform. While these methods can extract fault features to some extent, they are often computationally intensive, have high hardware requirements, and their stability and accuracy drop significantly in environments with low signal-to-noise ratios. Failure to accurately and robustly extract the core feature information reflecting the fault location will introduce significant errors into subsequent location calculations, delaying emergency repairs. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a 35kV power grid fault location method based on an improved Prony method, aiming to solve the problems in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a 35kV power grid fault location method based on the improved Prony method, comprising the following steps:

[0007] Step S1: Fault confirmation and line isolation; When a fault occurs in a 35kV power grid line, the protection device will activate and disconnect the circuit breakers at both ends of the line, so that the faulty line section is completely disconnected from the power grid operation state.

[0008] Step S2: Collaborative data acquisition at multiple measurement points at both ends of the power grid line; a transient excitation signal generated by capacitor discharge is injected into the power grid line through the fault location module at both ends of the power grid line, and the damped oscillating current signal generated in this process is synchronously recorded by multiple data acquisition units distributed at different locations of the power grid line to obtain two sets of multi-measurement point time series data.

[0009] Step S3: Feature parameter identification; The two sets of multi-point time series data are processed by the improved Prony algorithm combined with symplectic geometric mode decomposition (SGMD). The SGMD technology is used to reconstruct the phase space and denoise the signal, and the feature parameters corresponding to each set of time series data are identified, including the attenuation coefficient and angular frequency.

[0010] Step S4: Fault distance calculation and location output; Substitute the identified feature parameters into the physical model equations of the line fault to solve the problem, calculate the distance between the fault point and the measuring end of multiple measuring points, perform statistical correction on the results, and output the location result.

[0011] Further, the specific process of step S2 is as follows: Define the two ends of the power grid line as end 1 and end 2, and set up data acquisition units at multiple nodes of the power grid line; connect a fault location module to end 1 of the power grid line. The fault location module has a built-in capacitor of known capacity. First, charge the capacitor to a preset initial voltage, and then control the capacitor to discharge to the faulty power grid line to generate a transient excitation signal. At the same time, the transient decaying oscillating current signal generated during the discharge process is recorded synchronously by the Hall current sensor and the data acquisition unit to obtain the time series data of multiple measurement points at end 1 of the power grid line; keep the parameters of the fault location module unchanged, connect it to end 2 of the power grid line, and repeat the charging and discharging operation to synchronously collect the transient decaying oscillating current signal generated during the discharge process at end 2 of the power grid line to obtain the time series data of multiple measurement points at end 2 of the power grid line.

[0012] Further, the specific process of step S3 is as follows: the time series data of multiple measurement points at ends 1 and 2 of the power grid line are converted into discrete-time signals according to the sampling interval; the improved Prony algorithm combined with symplectic geometric mode decomposition (SGMD) is applied to the discrete-time signals: first, the phase space of the discrete signal is reconstructed to obtain the trajectory matrix, a symplectic matrix is ​​constructed and the symplectic eigenvalues ​​are solved, the effective signal components are determined according to the distribution characteristics of the symplectic eigenvalues, and the denoised signal sequence is reconstructed using symplectic geometric transformation and diagonal averaging; then, the order of the model is determined according to the number of effective symplectic eigenvalues, the denoised signal is fitted to the deterministic mathematical model, and the attenuation coefficient and angular frequency in the model are identified as characteristic parameters.

[0013] Furthermore, the specific process of substituting the identified characteristic parameters into the physical model equations of the line fault for solution is as follows:

[0014] Based on the damping characteristics of an RLC series circuit, the relationship between the total circuit resistance, attenuation coefficient, and angular frequency is defined.

[0015] For end 1 of the power grid line, the total circuit resistance is the line resistance from end 1 of the power grid line to the fault point plus the fault transition resistance; for end 2 of the power grid line, the total circuit resistance is the line resistance from end 2 of the power grid line to the fault point plus the fault transition resistance; combining the total line resistance, establish a system of equations; by subtracting the equations from the system of equations to eliminate the fault transition resistance, the line resistance from end 1 of the power grid line to the fault point and the line resistance from end 2 of the power grid line to the fault point can be obtained.

[0016] Furthermore, the specific process for calculating the distance from the fault point to the measuring end is as follows: Based on the unit length resistance of the power grid line, combined with the line resistance from end 1 of the power grid line to the fault point and the line resistance from end 2 of the power grid line to the fault point, multiple sets of distances from the fault point to end 1 of the line and distances from the fault point to end 2 of the line are calculated. Statistical corrections are then performed on the multiple sets of distances from the fault point to end 1 of the line and distances from the fault point to end 2 of the line to obtain the location result.

[0017] Furthermore, the preset initial voltage range is 10kV-35kV, and is dynamically adjusted according to the length of the faulty line.

[0018] Furthermore, before applying the improved Prony algorithm to the discrete-time signal, the discrete-time signal is preprocessed by a moving average filter; the window length of the moving average filter is preset to filter out high-frequency random noise in the signal.

[0019] An electronic device includes a processor, a memory, and a bus, wherein the processor and the memory are connected via the bus, wherein the memory is used to store a set of program code, and the processor is used to call the program code stored in the memory to execute a 35kV power grid fault location method based on the improved Prony method.

[0020] A non-volatile computer storage medium storing computer-executable instructions that execute a 35kV power grid fault location method based on an improved Prony method.

[0021] Compared with existing technologies, the present invention has the following advantages:

[0022] (1) This invention innovatively utilizes circuit breakers to isolate faulty lines and actively injects capacitor discharge signals to transform the complex fault network into a second-order RLC damped oscillating circuit with well-defined physical characteristics. By combining the two-end measurement data to construct a set of equations, the influence of unknown fault transition resistance on ranging accuracy is completely eliminated by the difference method, thus solving the core problem of large positioning error in the traditional impedance method when transition resistance exists.

[0023] (2) To address the problem of strong noise interference in power field applications, this invention introduces symplectic geometric mode decomposition (SGMD) technology. Utilizing Takens' embedding theorem, a one-dimensional time series is mapped to a high-dimensional phase space. By constructing a Hamiltonian matrix and performing a symplectic similarity transformation, the noise subspace and signal subspace are effectively separated while maintaining the signal topology. Compared to traditional filtering methods, SGMD can more accurately extract weak RLC oscillation features submerged in noise, significantly improving the parameter identification accuracy and robustness of the Prony algorithm in low signal-to-noise ratio environments.

[0024] (3) This invention overcomes the limitations of single-point measurement and constructs a multi-point collaborative measurement network. By synchronously collecting and analyzing data from multiple sensors distributed in different locations, and by introducing a statistical outlier removal and weighted average correction mechanism, it effectively avoids random errors caused by single measurement or single-point interference, and significantly improves the stability and reliability of fault location results.

[0025] (4) The present invention uses SGMD to accurately extract effective signal components, which can adaptively set the model order to a lower value, greatly reducing the computational complexity, eliminating the need to process massive transient data, and reducing hardware configuration requirements and costs; at the same time, the fast and accurate fault location can shorten the troubleshooting time, help the power sector to repair efficiently, reduce the power outage time, and ensure the stability of 35KV power grid operation and power supply reliability. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method of the present invention.

[0027] Figure 2 A comparison diagram of discharge current between the π-type equivalent model and the RL-type equivalent model.

[0028] Figure 3The diagram shows the improved performance of the Prony algorithm in recognizing capacitor discharge current.

[0029] Figure 4 The diagram shows the improved performance of the Prony algorithm in recognizing noisy signals. Detailed Implementation

[0030] like Figure 1 As shown, the present invention provides a technical solution: a 35kV power grid fault location method based on the improved Prony method, comprising the following steps:

[0031] Step S1: Fault confirmation and line isolation; When a fault occurs in a 35kV power grid line, the protection device will activate and disconnect the circuit breakers at both ends of the line, so that the faulty line section is removed from the power grid operation state.

[0032] Step S2: Collaborative data acquisition at multiple measurement points at both ends of the power grid line; a transient excitation signal generated by capacitor discharge is injected into the power grid line through the fault location module at both ends of the power grid line, and the damped oscillating current signal generated in this process is synchronously recorded by multiple data acquisition units distributed at different locations of the power grid line, so as to obtain two sets of multi-measurement point time series data.

[0033] Step S3: Feature parameter identification; The two sets of multi-point time series data are processed by the improved Prony algorithm combined with symplectic geometric mode decomposition (SGMD). The phase space is reconstructed and the signal is denoised using SGMD technology to identify the feature parameters corresponding to each set of time series data, including the attenuation coefficient and angular frequency.

[0034] Step S4: Fault distance calculation and location output; Substitute the identified characteristic parameters into the physical model equations of the line fault to solve the problem, calculate the distance from the fault point to the measuring end, and use the calculation results from multiple measuring points for statistical correction to output the location result.

[0035] The specific process of step S2 is as follows: define the two ends of the power grid line as end 1 and end 2, and set up data acquisition units at multiple key nodes of the power grid line (such as the beginning and end points, branch points, or sectionalizing switches); connect a fault location module to end 1 of the power grid line, and the fault location module has a built-in capacitor of known capacity. First Charge to the preset initial voltage Then control Discharging the faulty power grid line generates a transient excitation signal. Simultaneously, a high-precision Hall current sensor and a data acquisition unit synchronously record the transient decaying oscillating current signal generated during the discharge process, obtaining time series data of multiple measurement points at end 1 of the power grid line. Keeping the parameters of the fault location module unchanged, it is connected to end 2 of the power grid line, and the charging and discharging operations are repeated. The transient decaying oscillating current signal generated during the discharge process at end 2 of the power grid line is synchronously acquired, obtaining time series data of multiple measurement points at end 2 of the power grid line.

[0036] Among them, the preset initial voltage The value range is 10kV-35kV, and it is dynamically adjusted according to the length of the faulty line; when the line length is greater than 50km, The value is taken as 25kV-35kV; when the line length is less than or equal to 50km, The value is set between 10kV and 25kV to ensure that the attenuation and oscillation characteristics of the discharge signal are complete and identifiable.

[0037] The specific process of step S3 is as follows:

[0038] When a transient excitation signal is injected at end 1 of a power grid line, the first... The time series data of each measuring point are When a transient excitation signal is injected at both ends of the power grid line, the first The time series data of each measuring point are , , Indicates the number of measurement points. Indicates time;

[0039] Will , By sampling interval Convert to discrete-time signal , (Including noise) Indicates a discrete-time index;

[0040] Discrete-time signal for each measurement point The improved Prony algorithm combining symplectic geometric mode decomposition (SGMD) is applied respectively:

[0041] By utilizing the principle of symplectic geometric similarity transformation, a one-dimensional discrete-time signal is mapped to a high-dimensional phase space for analysis, thereby achieving the separation of noise and effective signal. First, phase space reconstruction is performed; for a length of... discrete-time signals Based on Takens' embedding theorem, the trajectory matrix is ​​constructed using the delayed coordinate method. :

[0042] ;

[0043] In the formula, Indicates the embedding dimension; Indicates a delayed signal; ;

[0044] Perform a symplectic geometric transformation based on the trajectory matrix. Constructing a symmetric matrix , express The transpose of the matrix is ​​used to construct the symplectic matrix. :

[0045] ;

[0046] In the formula, express Transpose of;

[0047] Perform symplectic eigenvalue solving and component reconstruction. For the symplectic matrix... Perform eigenvalue decomposition and solve the characteristic equation. , Represents the symplectic eigenvalue. The identity matrix is ​​represented by multiple symplectic eigenvalues; these symplectic eigenvalues ​​are the opposites of each other, and their moduli are taken and arranged in descending order.

[0048] Based on the distribution characteristics of symplectic eigenvalues ​​(the eigenvalues ​​corresponding to noise are usually small and randomly distributed, while the eigenvalues ​​corresponding to effective RLC damped oscillation signals are large and appear in pairs), a threshold for symplectic eigenvalues ​​is set or the number of dominant components is determined using the difference spectrum. ;

[0049] Before selection Construct the symplectic geometric transformation matrix for each symplectic eigenvalue. , Indicates the first The symplectic eigenvectors corresponding to the symplectic eigenvalues; for the trajectory matrix By performing mapping and reconstruction, the denoised reconstructed trajectory matrix is ​​obtained. , express Transpose of;

[0050] The trajectory matrix is ​​reconstructed using the diagonal average method. Inverse mapping back to a one-dimensional time series yields the denoised symplectic geometric mode components. :

[0051] ;

[0052] In the formula, Indicates the interval partitioning parameters; Represents the reconstructed trajectory matrix The Middle line, number The specific values ​​in the column; Represents the reconstructed trajectory matrix The Middle line, number The specific values ​​in the column; Represents the reconstructed trajectory matrix The Middle line, number The specific values ​​in the column;

[0053] Symmetric geometric mode components Substitute the data into the Prony algorithm, set the model order, and fit it to a deterministic mathematical model:

[0054] ;

[0055] ;

[0056] In the formula, , This represents the symplectic geometric mode component corresponding to the time series data when a transient excitation signal is injected at end 1 of the power grid line. This represents the symplectic geometric mode component corresponding to the time series data when a transient excitation signal is injected at both ends of the power grid line. express The model order represents the number of dominant decaying oscillation components in the signal; express The Middle The amplitude of each component; express The Middle The initial phase of each component; express The Middle The attenuation coefficient of each component; express The Middle The angular frequency of each component; Represents the natural constant; Represents the imaginary unit; express The model order represents the number of dominant decaying oscillation components in the signal; express The Middle The amplitude of each component; express The Middle The initial phase of each component; express The Middle The attenuation coefficient of each component; express The Middle The angular frequency of each component;

[0057] Will , , , As a feature parameter.

[0058] Before applying the improved Prony algorithm to the discrete-time signal, the discrete-time signal is preprocessed by a moving average filter. The window length of the moving average filter is 3-7 sampling intervals, which filters out high-frequency random noise in the signal and improves the stability of feature parameter identification.

[0059] When applying the improved Prony algorithm, the core parameter is the model order. The choice of model order. Given that fault location occurs after line isolation, the discharge circuit can be equivalent to a second-order RLC circuit with well-defined physical characteristics, whose transient current is mainly composed of a dominant damped oscillating component. Therefore, the model order... This can be set to a small value, which greatly reduces computational complexity and ensures algorithm stability. Furthermore, the frequency setting during discretization sampling must ensure that it can capture the data without distortion. and line inductance The highest oscillation frequency component is determined The window length for data analysis needs to be appropriate to fully encompass the signal's attenuation characteristics, thereby ensuring the accuracy of attenuation coefficient identification.

[0060] For a second-order RLC discharge circuit consisting of a fault location module, the differential equation of the circuit according to Kirchhoff's voltage law is:

[0061] ;

[0062] In the formula, Represents the equivalent inductance of a power grid line; Indicates the additional line inductance of the power grid line; express Discharge current at all times; This represents the line resistance from one end of the power grid line to the fault point. Represents the differential operator; This indicates the fault transition resistance.

[0063] The specific process of substituting the identified characteristic parameters into the physical model equations of the line fault for solution is as follows:

[0064] Based on the damping characteristics of an RLC series circuit, the total resistance of the loop is defined. With attenuation coefficient angular frequency The relationship is: ;

[0065] For terminal 1 of the power grid line, the total resistance of the loop is , This represents the line resistance from one end of the power grid line to the fault point. This represents the fault transition resistance; for terminals 2 of the power grid line, the total circuit resistance is... , This represents the line resistance from terminal 2 of the power grid to the fault point; combined with the total line resistance. Establish a system of equations:

[0066] ;

[0067] Eliminate by taking the difference of the system of equations (To avoid the impact of transition resistance uncertainty on fault location), the solution is obtained. and .

[0068] Among them, attenuation coefficient angular frequency Determined by the following formula:

[0069] ;

[0070] .

[0071] Regarding the feature parameters corresponding to time series data, the core lies in improving the Prony algorithm to process the collected time series data. , The damped sine wave mathematical model is expressed as follows:

[0072] ;

[0073] ;

[0074] In the formula, This represents the sine function.

[0075] The specific process for calculating the distance from the fault point to the measuring end is as follows:

[0076] according to The line resistance from one end of the power line at each measuring point to the fault point as well as The line resistance from terminal 2 of the power grid at each measuring point to the fault point and the resistance per unit length of the power grid line Substitute into the formula: , The distances from the fault point to end 1 of the line and the distances from the fault point to end 2 of the line were calculated from multiple measuring points. Indicates the first The distance from the fault point of each measuring point to end 1 of the line. Indicates the first The distance of the fault point of each measuring point from the two ends of the line.

[0077] Statistical methods were used to correct the distances from the fault point to end 1 and the fault point to end 2 of the line at multiple measurement points:

[0078] Taking the distance at one end of the line as an example, calculate arithmetic mean with standard deviation , is represented as:

[0079] ;

[0080] ;

[0081] Remove outliers that deviate significantly from the mean. Set a confidence interval threshold coefficient. (usually taken) ), retain the valid results that meet the conditions, and form a set of valid results. :

[0082] ;

[0083] A weighted average is calculated for the remaining valid results. Weighting coefficients are introduced. ( The final distance from the fault point to end 1 of the line can be determined based on the signal-to-noise ratio or fitting residual of each measuring point (the higher the signal-to-noise ratio, the greater the weight). :

[0084] ;

[0085] Similarly, the final distance from the fault point to both ends of the line can be calculated. This enables multi-point collaborative calibration, ensuring high robustness of the positioning results.

[0086] Among them, the resistance per unit length The resistance is obtained through real-time measurement. Before fault location, a calibration signal is injected into the normally operating power grid line through the fault location module, the response data is collected to calculate the actual resistance per unit length of the line, and the preset fixed value is replaced to eliminate the influence of line temperature changes and aging on the resistance parameters.

[0087] When the improved Prony algorithm is applied to fault location, the identified attenuation coefficient and angular frequency directly correspond to the total resistance and total reactance of the fault circuit, possessing clear physical meaning. These two sets of parameters, obtained through two-end measurements, can be directly substituted into the physical equations for solution, effectively eliminating the influence of uncertainties such as transition resistance, ultimately achieving high-precision fault distance calculation. Unlike the frequency amplitude extracted by methods such as STFT, the attenuation coefficient and angular frequency directly correspond to the total resistance and total reactance of the fault circuit, possessing clear physical meaning. These two sets of parameters, obtained through two-end measurements, can be directly substituted into the physical equations for solution, effectively eliminating the influence of uncertainties such as transition resistance, ultimately achieving high-precision fault distance calculation.

[0088] An electronic device includes a processor, a memory, and a bus, wherein the processor and the memory are connected via the bus, wherein the memory is used to store a set of program code, and the processor is used to call the program code stored in the memory to execute a 35kV power grid fault location method based on the improved Prony method.

[0089] A non-volatile computer storage medium storing computer-executable instructions that execute a 35kV power grid fault location method based on an improved Prony method.

[0090] like Figure 2 As shown, Figure 2 In the figure, (a) represents the discharge current when using the π-type equivalent model. Figure 2 In the figure, (b) represents the discharge current when using the RL-type equivalent model. Figure 2 (c) represents the difference between the two. In power system analysis, the π-type model is often used to characterize the distributed parameter characteristics of actual transmission lines. In this invention, after the faulty line is isolated in step S1, the line itself can be simplified and equivalent to an RL-type model (i.e., only its resistance and inductance characteristics are considered); at this time, since the fault location module with built-in capacitor is connected in step S2, this RL-type equivalent model together with the location device constitutes a second-order RLC discharge circuit with clearly defined physical characteristics. Figure 1 Simulation results validated the feasibility of this simplification: the response waveform of the RLC discharge circuit constructed using the simplified RL-type circuit model is highly consistent with the waveform trend of the RLC circuit constructed using the complex π-type circuit model, and the difference between the two is close to 0. This proves that analyzing the "isolated faulty circuit" and the "fault location module" together as equivalent to a second-order RLC discharge circuit has high accuracy.

[0091] like Figure 3 As shown, Figure 3 In the middle (a), the original discharge current signal is shown. Figure 3In the middle (b), the signal obtained by fitting the original Prony algorithm is shown. Figure 3 (c) shows the signal fitted by the improved Prony algorithm. The fitting effects of the original Prony algorithm and the improved Prony algorithm were compared under an ideal noise-free environment. It can be observed that both can reproduce the waveform of the original discharge current signal well. In particular, the waveform of the signal fitted by the improved Prony algorithm almost completely overlaps with the original discharge current signal, proving that in the absence of noise interference, the improved Prony algorithm can accurately reconstruct the effective signal, preserving core features such as the attenuation coefficient and angular frequency, providing reliable parameter input for high-precision fault location.

[0092] like Figure 4 As shown, Figure 4 In the middle (a), the initial signal contains 30dB of noise. Figure 4 (b) shows the signal obtained by the original Prony algorithm. Figure 4 (c) shows the signal identified by the improved Prony algorithm. In a harsh environment with 30dB of strong noise, the original Prony algorithm, while filtering out some noise, still exhibits significant glitches and waveform distortion, making it difficult to accurately pinpoint signal features. In contrast, the improved Prony algorithm successfully filters out most of the random noise, recovering a smooth, highly consistent decaying oscillation characteristic with the pure signal. This strongly demonstrates the superiority of the improved Prony algorithm based on phase space reconstruction, enabling stable extraction of physical parameters even in low signal-to-noise ratio environments, significantly ensuring the reliability of fault location results.

[0093] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for fault location in 35 KV power grid based on improved Prony method, characterized in that, Comprising the following steps: Step S1: fault confirmation and line isolation; when a fault occurs in a 35kV power grid line, a protection device acts to disconnect circuit breakers at both ends of the line, so that the fault line section is completely disconnected from the power grid operation state; Step S2: dual-end multi-measurement-point cooperative data acquisition; at both ends of the power grid line, a transient excitation signal generated by capacitor discharge is injected into the power grid line through a fault location module, and the decaying oscillation current signal generated in this process is recorded synchronously by multiple data acquisition units distributed at different positions of the power grid line, thereby obtaining two sets of multi-measurement-point time series data; Step S3: feature parameter identification; The two sets of multi-measurement-point time series data obtained are respectively processed by applying an improved Prony algorithm combined with symplectic geometric modal decomposition (SGMD), the SGMD technology is used for phase space reconstruction and signal denoising, and the feature parameters corresponding to each set of time series data are identified, including the attenuation coefficient and the angular frequency; Step S4: fault distance calculation and positioning output; The identified feature parameters are substituted into the physical model equation set of the line fault to solve, the distance results of the fault point from the measurement end of multiple measurement points are calculated, the results are statistically corrected, and the positioning results are output; The specific process of step S2 is as follows: define the two ends of the power grid line as 1 end and 2 end, and set data acquisition units at multiple nodes of the power grid line; connect the fault location module to the 1 end of the power grid line, the fault location module is built-in a capacitor with a known capacity, the capacitor is first charged to a preset initial voltage, and then the capacitor is controlled to discharge to generate a transient excitation signal to the fault power grid line, at the same time, the transient decaying oscillation current signal generated in the discharging process is recorded synchronously by a Hall current sensor and a data acquisition unit, thereby obtaining multi-measurement-point time series data at the 1 end of the power grid line; The parameters of the fault location module remain unchanged, the fault location module is connected to the 2 end of the power grid line, the charging and discharging operations are repeated, the transient decaying oscillation current signal generated in the discharging process at the 2 end of the power grid line is synchronously acquired, and multi-measurement-point time series data at the 2 end of the power grid line are obtained; The specific process of step S3 is as follows: the multi-measurement-point time series data at the 1 end and the 2 end of the power grid line are converted into discrete time signals according to the sampling interval; the improved Prony algorithm combined with the symplectic geometric modal decomposition (SGMD) is applied to the discrete time signals: firstly, the trajectory matrix is obtained by phase space reconstruction of the discrete signal, the symplectic matrix is constructed and the symplectic eigenvalue is solved, the effective signal component is determined according to the distribution characteristics of the symplectic eigenvalue, and the signal sequence after denoising is reconstructed by using the symplectic geometric transformation and the diagonal average method; then the order of the model is determined according to the number of effective symplectic eigenvalues, the signal after denoising is fitted to the determined mathematical model, and the attenuation coefficient and the angular frequency in the model are identified as the feature parameters.

2. The 35 KV power grid fault location method based on improved Prony method according to claim 1, characterized in that: The specific process of substituting the identified feature parameters into the physical model equation set of the line fault to solve is as follows: According to the damping characteristics of the RLC series circuit, the relationship between the total loop resistance and the attenuation coefficient and the angular frequency is defined; The total loop resistance at the power grid line 1 end is the line resistance from the power grid line 1 end to the fault point plus the fault transition resistance; the total loop resistance at the power grid line 2 end is the line resistance from the power grid line 2 end to the fault point plus the fault transition resistance; an equation set is established in combination with the line total resistance; the fault transition resistance is eliminated by difference through the equation set, and the line resistance from the power grid line 1 end to the fault point and the line resistance from the power grid line 2 end to the fault point are solved.

3. The 35 KV power grid fault location method based on improved Prony method according to claim 2, characterized in that: The specific process of calculating the distance of the fault point from the measuring end is as follows: according to the unit length resistance of the power grid line, in combination with the line resistance from the power grid line 1 end to the fault point and the line resistance from the power grid line 2 end to the fault point, a plurality of sets of results of the distance of the fault point from the power grid line 1 end and the distance of the fault point from the power grid line 2 end are calculated, the results of the distance of the fault point from the power grid line 1 end and the distance of the fault point from the power grid line 2 end are statistically corrected, and a positioning result is obtained.

4. The 35 KV power grid fault location method based on improved Prony method according to claim 3, characterized in that: The preset initial voltage is in a range of 10kV-35kV, and is dynamically adjusted according to the length of the fault line.

5. The 35 KV power grid fault location method based on improved Prony method according to claim 4, characterized in that: Before the improved Prony algorithm is applied to the discrete time signal, the discrete time signal is preprocessed by sliding average filtering; a window length of the sliding average filtering is preset, and high-frequency random noise in the signal is filtered out.

6. An electronic device, comprising: The computer executable instructions perform the 35KV power grid fault positioning method based on the improved Prony method in any one of claims 1-5.

7. A non-transitory computer storage medium storing computer-executable instructions, the computer-executable instructions comprising instructions for: The computer executable instructions perform the 35KV power grid fault positioning method based on the improved Prony method in any one of claims 1-5.

Images