Single-end distance measurement method and system based on time-frequency transformation and forward and backward fault traveling wave fusion

Abstract

The invention discloses a single-ended distance measurement method and system based on time-frequency transformation and forward and backward fault traveling wave fusion, by introducing synchronous extrusion transformation (SST), the high time resolution is maintained, the frequency resolution is remarkably improved, the transient characteristics of fault traveling waves are more accurately captured, and the fault traveling waves can be accurately measured by combining the time-frequency characteristics of the forward and backward fault traveling waves. According to the method, a comprehensive distance measurement function is constructed, the precision and robustness of fault positioning are further improved, meanwhile, the time window width is dynamically optimized by adopting an adaptive adjustment strategy, the calculation complexity is reduced, and the real-time performance and efficiency of an algorithm are improved. According to the invention, a more efficient and reliable technical means is provided for fault detection and positioning of the power system.

Description

Technical Field

[0001] This invention relates to the field of power system fault location, specifically to a single-end ranging method and system based on time-frequency transformation and forward / backward fault traveling wave fusion. Background Technology

[0002] In modern power systems, the reliability and safety of transmission lines are of paramount importance. With the continuous expansion and increasing complexity of power grids, fault detection and location have become crucial for ensuring the stable operation of power systems. Traditional fault location methods primarily rely on the measurement of voltage and current signals and analysis using various algorithms. However, these traditional methods face numerous challenges in practical applications.

[0003] First, traveling wave signals in power systems exhibit non-stationary characteristics, meaning their frequency components vary over time. Traditional Fourier transforms cannot effectively capture this time-varying characteristic, resulting in insufficient fault location accuracy. Second, power systems contain numerous noise and interference sources, such as electromagnetic interference and harmonics. These factors significantly degrade the quality of fault signals, making it difficult to accurately identify fault locations using methods based on simple time-domain or frequency-domain analysis.

[0004] In recent years, time-frequency transformation technology has been increasingly applied to fault detection and location in power systems. Time-frequency transformation provides a joint representation of a signal in both time and frequency, thus better capturing the characteristics of non-stationary signals. Among these methods, short-time Fourier transform, wavelet transform, and synchronous squeezing transform have been widely used. These methods have improved the accuracy of fault location to some extent, but still have some limitations. For example, there is a trade-off between the time resolution and frequency resolution of short-time Fourier transform, wavelet transform requires the selection of appropriate basis functions, and while synchronous squeezing transform can improve time-frequency resolution, it still needs further optimization when dealing with complex and variable fault modes. Meanwhile, the fusion of forward and backward traveling waves is also an important means to improve fault location accuracy. Forward traveling waves propagate from the power source to the load end, generating reverse traveling waves upon encountering a fault point. By comprehensively analyzing the forward and reverse traveling waves, the fault location can be determined more accurately. However, existing schemes for fusing forward and backward traveling waves mostly focus on simple time-domain or frequency-domain analysis, failing to fully utilize the advantages of time-frequency transformation, resulting in room for improvement in fault location accuracy.

[0005] In summary, existing fault location methods have certain limitations when dealing with complex power systems, and there is an urgent need for a more efficient and accurate single-end ranging scheme to meet practical needs. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a single-end ranging method based on the fusion of time-frequency transformation and forward and backward fault traveling waves to address the shortcomings of existing methods and solve the common-mode voltage problem caused by high-frequency injection in modular multilevel converters.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A single-end ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion includes the following steps: Step 1: Collect the traveling wave signal of the fault voltage of the transmission line when a power system fault occurs; Step 2: By iterating through the signal sample points of the fault voltage traveling wave signal using a sliding time window, a series of weighted variance values ​​that vary with time are obtained. ,when rate of change At that time, the voltage traveling wave signal within the corresponding sliding time window is identified as the point of change of fault characteristics; Step 3: Perform continuous wavelet transform on the voltage traveling wave signal at the fault characteristic change point, and select Morlet wavelet to calculate the instantaneous frequency and redistribute the energy to obtain the time-frequency representation of the fault voltage traveling wave signal. Step 4: Construct a time-frequency diagram based on the time-frequency representation; Step 5: Determine the low-frequency and high-frequency bands of the time-frequency diagram, and calculate the energy proportion of the low-frequency band. and the proportion of high-frequency energy And set the energy percentage threshold. and ;when When it is determined to be a forward traveling wave of fault voltage, when The time is judged as a backward traveling wave of fault voltage; if it appears and The priority is determined by the extent to which the proportion exceeds the threshold: calculate the high-frequency proportion excess rate. and low-frequency excess rate ,like If it is determined to be a backward traveling wave, then it is determined to be a forward traveling wave. Step 6: Determine the time it takes for the forward or backward traveling wave of the fault voltage to reach one end of the measurement point, calculate the time difference, and calculate the fault distance based on the traveling wave velocity.

[0008] A further improvement is made to the second step, where the window length of the sliding time window is... Sliding step size ;in, It is a positive integer and its initial value is between 25 and 50. It is a positive integer, initially ranging from 5 to 10. The sampling period; when the rate of change Three consecutive windows exceed the first threshold ,but , When the rate of change Three consecutive windows below the second threshold ,but , .

[0009] In a further improvement, in step two, ; in, It is a discrete fault voltage traveling wave signal sequence. Represents the first in the sequence One sample point, It is the total number of sample points within the sliding time window. It is the mean of the traveling wave signal within the sliding time window, calculated using the following formula: , The sample weights are calculated using the following formula: .

[0010] A further improvement is made in step two, where the rate of change... , This represents the weighted variance value of the current nth window. This is the weighted variance value of the previous window.

[0011] Further improvements are made, and the specific steps of step three are as follows: Traveling wave signal of fault voltage Perform continuous wavelet transform: ; Represents the wavelet coefficients at different scales and time points, where It is a scaling parameter that controls the time scale of the wavelet; larger values ​​correspond to lower frequencies, and smaller values ​​correspond to higher frequencies. It is a translation parameter that controls the position of the wavelet on the time axis, corresponding to the time point of the signal; Represents the complex conjugate of wavelet basis functions; It is a wavelet basis function, and Morlet wavelet is selected; Represents the integral variable, i.e., continuous time; As a normalization factor, it ensures the energy consistency of wavelets at different scales and avoids energy distortion caused by scale changes; The Morlet wavelet basis function expression is: , Let be the oscillation frequency of the wavelet in the time domain, and take the center frequency as . , The imaginary unit satisfies where e is the natural logarithm, and then the scale is calculated. and time instantaneous frequency at point ,in Represents the imaginary part of a complex number. wavelet coefficients For translation parameters The partial derivatives, corresponding to the time derivatives, reflect the rate of change of phase with time; based on this process, the wavelet transform coefficients are transformed from the scale plane... Transform to frequency plane Finally, a synchronous extrusion transformation is performed: For each target frequency traverse all scales Find the corresponding instantaneous frequency and using the Dirac function Redistribute energy to the target frequency superior; This is a time-frequency representation.

[0012] Further improvements are made to step four, which is as follows: Based on time-frequency representation Construct a two-dimensional time-frequency graph, where the horizontal axis represents time, the vertical axis represents frequency, and the color represents energy intensity.

[0013] As a further improvement, step five is as follows: Calculate the energy proportion of the low-frequency band and the energy proportion of the high-frequency band: ; Integration interval and Selected based on the low-frequency and high-frequency ranges of the time-frequency diagram. These represent the lower and upper limits of the entire frequency range of the synchronous extrusion transformation results, respectively. Integrating along the time axis to obtain the global frequency energy spectrum describing the overall distribution of signal energy in the frequency dimension, the frequency corresponding to the peak value is identified and denoted as the dominant frequency. , It characterizes the frequency components with the most concentrated and significant energy in the current fault traveling wave signal; the low-frequency integration interval Defined as The proportionality coefficient A positive real number less than 1, with a value of 0.5, to ensure that the low-frequency band covers the main low-frequency components from the lowest analysis frequency to below the dominant frequency; the high-frequency band integration interval Defined as The proportionality coefficient It is a positive real number greater than 1, with a value of 2.0, to ensure effective separation of the main high-frequency components that are higher than the dominant frequency and excited by the sudden change at the fault point.

[0014] Further improvements, energy percentage threshold and The determination is made dynamically by combining the current operating status of the power grid and historical fault data using machine learning methods, as follows: a) Data preparation: Collect samples containing power grid operating status parameters and pre- and post-fault traveling wave frequency components; the operating status parameters include voltage fluctuation range. Current distortion rate and power factor The forward and backward traveling wave frequency component data of the fault includes the proportion of low-frequency energy. and the proportion of high-frequency energy And mark the forward traveling wave time and backward traveling wave , The samples are labeled; then the training set, validation set, and test set are divided into 70%, 15%, and 15% respectively; the data of the samples are uniformly normalized. b) Model Training: An SVM model with a radial basis function kernel is selected. The Gaussian kernel function expression of the SVM model is as follows: , in, Indicates the first The feature vector of each sample Indicates the first The feature vector of each sample Represents the Gaussian kernel function. Indicates bandwidth parameter; Set initial hyperparameters: penalty coefficient C = 1.0, bandwidth parameter During the training process, the model parameters are determined by solving the following dual optimization problem of SVM: Maximize the objective function , in It is the number of training samples. and It is the first The and the first The labels of each sample simultaneously satisfy the constraints. , ; For the first Lagrange multipliers for a sample, Indicates the first Lagrange multipliers for each sample; During training, the Sequence Minimum Optimization (SMO) algorithm is used to solve the optimization problem of maximizing the objective function. c) Model Evaluation and Parameter Tuning: Input the validation set data into the trained SVM model, calculate the evaluation metrics of the trained SVM model, including prediction accuracy, recall, and F1 score. Based on the evaluation results of the validation set, adjust the hyperparameters of the SVM model using grid search. and Find the optimal combination of hyperparameters; d) Threshold determination: Using the trained model to predict on the test set, the average energy proportions of the low-frequency and high-frequency bands in the forward and backward traveling wave samples of the fault are statistically analyzed and denoted as... ;Pick and The average value is used as the threshold for the proportion of low-frequency energy. ,Pick and The average value is used as the threshold for the proportion of energy in the high-frequency band. .

[0015] Further improvements, and Dynamic updates are performed: Based on the real-time power grid status and newly added fault data, the SVM model is retrained after every 100 sets of fault data are added.

[0016] Further improvements include the following specific steps in step six: determining the arrival times of the forward and backward traveling waves at measurement point M. and Calculate the time difference Given the propagation speed of traveling waves in a DC power grid According to the formula Calculate the distance to the fault and determine the fault point F.

[0017] A single-ended ranging system based on time-frequency transformation and forward / backward fault traveling wave fusion is disclosed. The system is used to run the aforementioned single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion. Compared with the prior art, the advantages of this invention are as follows: 1. Improved Time-Frequency Resolution: By introducing Synchronous Squeeze Transform (SST), its advantages in time-frequency resolution are fully utilized, overcoming the limitations of traditional STFT and wavelet transform. Synchronous Squeeze Transform can significantly improve frequency resolution while maintaining high time resolution, thereby more accurately capturing the transient characteristics of fault traveling waves.

[0018] 2. Optimized Forward and Backward Traveling Wave Fusion: A comprehensive ranging function is constructed by combining the time-frequency characteristics of forward and backward fault traveling waves. Through correlation analysis and feature extraction, the similarities and differences between the forward and reverse traveling waves of the fault voltage are fully utilized to achieve more accurate fault location. This method not only considers the time difference but also integrates factors such as waveform morphology, enhancing the robustness and adaptability of the algorithm.

[0019] 3. Reduce computational complexity: To address the high computational complexity of synchronous extrusion transformation, an adaptive adjustment strategy is proposed. The time window width is dynamically adjusted based on the changing trend of the correlation coefficient, reducing unnecessary computation and improving the algorithm's real-time performance and efficiency. Furthermore, multi-scale analysis and optimized parameter settings further reduce the computational burden.

[0020] 4. Enhanced anti-interference capability: In the signal preprocessing stage, advanced denoising technology is employed to effectively remove noise and interference while retaining useful fault characteristic information. By combining multi-source data and intelligent algorithms, the system's anti-interference capability and stability are improved. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of a single-end ranging scheme based on time-frequency transformation and forward / backward fault traveling wave fusion according to the present invention. Figure 2 This is a schematic diagram of a single-ended ranging scheme based on time-frequency transformation and forward / backward fault traveling wave fusion to simulate a fault occurrence according to the present invention. Figure 3 This is a functional module diagram of a single-end ranging scheme based on time-frequency transformation and forward / backward fault traveling wave fusion according to the present invention.

[0022] Figure 4 This is the present invention. Figure 1 A detailed flowchart illustrating the process of refining the energy percentage threshold in step 5.

[0023] Figure 5 This is the present invention. Figure 1 A schematic diagram of the time-frequency diagram of synchronous extrusion transformation in step 4. Detailed Implementation

[0024] A single-end ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion includes the following steps:

[0025] Data Acquisition: Using high sampling rate data acquisition devices, the traveling wave signal of the transmission line fault voltage at the time of the fault is acquired from the power system.

[0026] Preprocessing: Assume the sampling period is... Window length It is an integer multiple of the sampling period, i.e. ,in It is a positive integer and its initial value is between 25 and 50. Sliding step size. Also set based on the sampling period, at the same time , are positive integers and The initial value is 5-10. As the time window slides gradually on the time axis according to the set sliding step size, the signal sample points within each time window are traversed, resulting in a series of weighted variance values ​​that change over time. By observing the trends in these variance values, a sudden increase in variance indicates a sharp increase in the dispersion of the signal within that time window, which is a characteristic change point related to the fault. When calculating the variance, a weighted variance calculation method is used, and the calculation formula is as follows: ,in It is a discrete signal sequence. Represents the first in the sequence One sample point, It is the number of sample points within the window. It is the mean of the signal within the window, calculated using the following formula: , The sample weights are calculated using the following formula: The weighting is based on the signal's distribution characteristics within the window. Sample points closer to the window center are more critical in reflecting the overall signal characteristics; therefore, sample points closer to the window center have greater weight. An adaptive adjustment mechanism is also introduced to automatically adjust the weighting based on the rate of change of the signal variance. and Value. When the rate of change of variance Three consecutive windows exceed the threshold ,but , When the rate of change of variance Three consecutive windows below the threshold ,but , .

[0027] Synchronous squeezing transform: By "squeezing" the wavelet transform coefficients—that is, redistributing energy to the most probable frequency positions—frequency resolution is improved while maintaining good time resolution. This involves the voltage traveling wave signal at these fault-related characteristic change points. Perform synchronous compression transform to obtain the time-frequency representation of the traveling wave signal. The calculation process is as follows: First, the voltage traveling wave signal... Perform continuous wavelet transform , representing wavelet coefficients at different scales and time points, where It is a scaling parameter that controls the time scale of the wavelet; larger values ​​correspond to lower frequencies, and smaller values ​​correspond to higher frequencies. These are translation parameters that control the position of the wavelet on the time axis; Represents the complex conjugate of wavelet basis functions; The wavelet basis function, Morlet wavelet, is selected. It is effective in distinguishing and extracting different frequency components in fault traveling wave signals, especially when analyzing the energy proportions of high and low frequencies, enabling a more accurate determination of the signal's energy distribution across different frequency ranges. In the frequency component analysis step of this scheme, Morlet wavelet helps to more accurately identify forward and backward traveling waves, improving the accuracy of ranging. Its expression is: , The center frequency is then used. The scale is then calculated. and time instantaneous frequency at point ,in and Let these represent the imaginary and real parts of the complex number, respectively. Based on this process, the wavelet transform coefficients are transformed from the scale plane... Transform to frequency plane Finally, a synchronous extrusion transformation is performed. For each target frequency traverse all scales Find the corresponding instantaneous frequency and using Dirac The function redistributes energy to the target frequency. Above, including Dirac The function is used to concentrate energy at the estimated instantaneous frequency, improve time-frequency resolution, reduce spectral aliasing, and thus more clearly show the time-frequency structure of the signal.

[0028] Time-frequency diagram generation: The time-frequency representation obtained from the above steps A two-dimensional time-frequency plot is constructed, with the horizontal axis representing time, the vertical axis representing frequency, and the color representing energy intensity. This time-frequency plot can intuitively display the energy distribution characteristics of the traveling wave signal at different times and frequencies.

[0029] Frequency component analysis: Calculating the energy proportion in the low-frequency band and the proportion of high-frequency energy When, the integration interval and Selected based on the low-frequency and high-frequency ranges of the time-frequency diagram. This covers the entire frequency range of the synchronous extrusion transformation results. Simultaneously, an energy percentage threshold is set. and ,when When it is determined to be a forward traveling wave of fault voltage, when The fault voltage is identified as a backward traveling wave, and the energy proportion threshold is dynamically determined using machine learning methods (support vector machine) combined with the current operating status of the power grid and historical fault data. The specific steps are as follows: a) Data preparation: Collect data including power grid operating status parameters (voltage fluctuation range) Current distortion rate Power factor ) and forward and backward traveling wave frequency component data (low-frequency energy proportion) High-frequency energy ratio Samples of forward traveling waves ( ) were labeled. ) and backward traveling wave ( Then, the dataset is divided into training, validation, and test sets, with each set comprising 70%, 15%, and 15% respectively. All data are then normalized. ( refer to , , , , ).

[0030] b) Model Training: An SVM model with a Gaussian kernel function is selected. The expression for the Gaussian kernel function is as follows: The optimized kernel function is a radial basis function (RBF), with a penalty coefficient C=1.0 and a bandwidth parameter. During training, the model parameters (Lagrange multipliers) are determined by solving the following optimization problem. ): Maximize the objective function as ,in It is the number of training samples. and The labels of the samples must also satisfy the constraints. , During training, the Sequence Minimum Optimization (SMO) algorithm is used to solve the above optimization problem.

[0031] c) Model Evaluation and Parameter Tuning: Input the validation set data into the trained SVM model and calculate evaluation metrics such as prediction accuracy, recall, and F1 score. Based on the evaluation results of the validation set, adjust the hyperparameters of the SVM model using grid search. and Find the optimal combination of hyperparameters.

[0032] d) Threshold determination: Using the trained model to predict on the test set, the average energy proportions of the low-frequency and high-frequency bands in the forward and backward traveling wave samples of the fault are statistically analyzed and denoted as... .Pick and The average value is used as the threshold for the proportion of low-frequency energy. ,Pick and The average value is used as the threshold for the proportion of energy in the high-frequency band. .

[0033] e) Threshold dynamic update: Based on the real-time power grid status and newly added fault data, the model is retrained after every 100 new sets of fault data.

[0034] Time difference calculation and ranging: Determining the arrival time of the forward and backward traveling waves of the fault at the measurement point M. and Calculate the time difference The propagation speed of a traveling wave in a DC power grid is known. According to the formula Calculate the distance to the fault and determine the fault point F.

[0035] The examples described above are only used to illustrate the technical solutions of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A single-end ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion, characterized in that, Includes the following steps: Step 1: Collect the traveling wave signal of the fault voltage of the transmission line when a power system fault occurs; Step 2: By iterating through the signal sample points of the fault voltage traveling wave signal using a sliding time window, a series of weighted variance values ​​that vary with time are obtained. ,when rate of change At that time, the voltage traveling wave signal within the corresponding sliding time window is identified as the point of change of fault characteristics; Step 3: Perform continuous wavelet transform on the voltage traveling wave signal at the fault characteristic change point, and at the same time select Morlet wavelet, calculate the instantaneous frequency and redistribute the energy to obtain the time-frequency representation of the fault voltage traveling wave signal; Step 4: Construct a time-frequency diagram based on the time-frequency representation; Step 5: Determine the low-frequency and high-frequency bands of the time-frequency diagram, and calculate the energy proportion of the low-frequency band. and the proportion of high-frequency energy And set the energy percentage threshold. and ;when When it is determined to be a forward traveling wave of fault voltage, when The fault voltage was identified as a backward traveling wave. If it appears and The priority is determined by the extent to which the proportion exceeds the threshold: calculate the high-frequency proportion excess rate. and low-frequency excess rate ,like If it is determined to be a backward traveling wave, then it is determined to be a forward traveling wave. Step 6: Determine the time it takes for the forward or backward traveling wave of the fault voltage to reach one end of the measurement point, calculate the time difference, and calculate the fault distance based on the traveling wave velocity.

2. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 1, characterized in that, In step two, the window length of the sliding time window Sliding step size ;in, It is a positive integer and its initial value is between 25 and 50. It is a positive integer, initially ranging from 5 to 10. The sampling period; when the rate of change Three consecutive windows exceed the first threshold ,but , When the rate of change Three consecutive windows below the second threshold ,but , .

3. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 1, characterized in that, In step two, ； in, It is a discrete fault voltage traveling wave signal sequence. Represents the first in the sequence One sample point, It is the total number of sample points within the sliding time window. It is the mean of the traveling wave signal within the sliding time window, calculated using the following formula: , The sample weights are calculated using the following formula: .

4. The single-end ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 1, characterized in that, In step two, the rate of change , This is the weighted variance value of the current nth window. This is the weighted variance value of the previous window.

5. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 1, characterized in that, The specific steps of step three are as follows: Traveling wave signal of fault voltage Perform continuous wavelet transform: ； Represents the wavelet coefficients at different scales and time points, where It is a scaling parameter that controls the time scale of the wavelet; larger values ​​correspond to lower frequencies, and smaller values ​​correspond to higher frequencies. It is a translation parameter that controls the position of the wavelet on the time axis, corresponding to the time point of the signal; Represents the complex conjugate of wavelet basis functions; It is a wavelet basis function, and Morlet wavelet is selected; Represents the integral variable, i.e., continuous time; As a normalization factor, it ensures the energy consistency of wavelets at different scales and avoids energy distortion caused by scale changes; The Morlet wavelet basis function expression is: , Let be the oscillation frequency of the wavelet in the time domain, and take the center frequency as . , The imaginary unit satisfies where e is the natural logarithm, and then the scale is calculated. and time instantaneous frequency at point ,in Represents the imaginary part of a complex number. Wavelet coefficients For translation parameters The partial derivative of , corresponding to the time derivative, reflects the rate of change of phase over time; Based on this process, the wavelet transform coefficients are transformed from the scale plane. Transform to frequency plane Finally, a synchronous extrusion transformation is performed: For each target frequency traverse all scales Find the corresponding instantaneous frequency and using the Dirac function Redistribute energy to the target frequency superior; This is a time-frequency representation.

6. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 5, characterized in that, The steps in step four are as follows: Based on time-frequency representation Construct a two-dimensional time-frequency graph, where the horizontal axis represents time, the vertical axis represents frequency, and the color represents energy intensity.

7. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 1, characterized in that, The method for calculating the low-frequency energy ratio and the high-frequency energy ratio in step five is as follows: ； Integration interval and Selected based on the low-frequency and high-frequency ranges of the time-frequency diagram. These represent the lower and upper limits of the entire frequency range of the synchronous extrusion transformation results, respectively. Integrating along the time axis to obtain the global frequency energy spectrum describing the overall distribution of signal energy in the frequency dimension, the frequency corresponding to the peak value is identified and denoted as the dominant frequency. , It characterizes the frequency components with the most concentrated and significant energy in the current fault traveling wave signal; the low-frequency integration interval Defined as The proportionality coefficient A positive real number less than 1, with a value of 0.5, to ensure that the low-frequency band covers the main low-frequency components from the lowest analysis frequency to below the dominant frequency; the high-frequency band integration interval Defined as The proportionality coefficient It is a positive real number greater than 1, with a value of 2.0, to ensure effective separation of the main high-frequency components that are higher than the dominant frequency and excited by the sudden change at the fault point.

8. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 7, characterized in that, Energy percentage threshold and The determination is made dynamically by combining the current operating status of the power grid and historical fault data using machine learning methods, as follows: a) Data preparation: Collect samples containing power grid operating status parameters and pre- and post-fault traveling wave frequency components; the operating status parameters include voltage fluctuation range. Current distortion rate and power factor The forward and backward traveling wave frequency component data of the fault includes the proportion of low-frequency energy. and the proportion of high-frequency energy And mark the forward traveling wave time and backward traveling wave , The samples are labeled; then the training set, validation set, and test set are divided into 70%, 15%, and 15% respectively; the data of the samples are uniformly normalized. b) Model Training: An SVM model with a radial basis function kernel is selected. The Gaussian kernel function expression of the SVM model is as follows: , in, Indicates the first The feature vector of each sample Indicates the first The feature vector of each sample Represents the Gaussian kernel function. Indicates bandwidth parameter; Set initial hyperparameters: penalty coefficient C = 1.0, bandwidth parameter During the training process, the model parameters are determined by solving the dual optimization problem of the SVM: Maximize the objective function , in It is the number of training samples. and It is the first The and the first The labels of each sample simultaneously satisfy the constraints. , ; For the first Lagrange multipliers for a sample, Indicates the first Lagrange multipliers for each sample; During training, the Sequence Minimum Optimization (SMO) algorithm is used to solve the optimization problem of maximizing the objective function. c) Model Evaluation and Parameter Tuning: Input the validation set data into the trained SVM model, calculate the evaluation metrics of the trained SVM model, including prediction accuracy, recall, and F1 score. Based on the evaluation results of the validation set, adjust the hyperparameters of the SVM model using grid search. and Find the optimal combination of hyperparameters; d) Threshold determination: Using the trained model to predict on the test set, the average energy proportions of the low-frequency and high-frequency bands in the forward and backward traveling wave samples of the fault are statistically analyzed and denoted as... ;Pick and The average value is used as the threshold for the proportion of low-frequency energy. ,Pick and The average value is used as the threshold for the proportion of energy in the high-frequency band. ; and Dynamic updates are performed: Based on the real-time power grid status and newly added fault data, the SVM model is retrained after every 100 sets of fault data are added.

9. The single-ended ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in claim 7, characterized in that, The specific steps of step six are as follows: Determine the arrival times of the forward and backward traveling waves of the fault at the measurement point M. and Calculate the time difference Given the propagation speed of traveling waves in a DC power grid According to the formula Calculate the distance to the fault and determine the fault point F.

10. A single-ended ranging system based on time-frequency transformation and forward / backward fault traveling wave fusion, characterized in that, The system is used to run the single-end ranging method based on time-frequency transformation and forward / backward fault traveling wave fusion as described in any one of claims 1-9.

Images