Railway catenary measurement method and system based on vibration compensation

Abstract

This invention belongs to the technical field of railway maintenance equipment, and relates to a measurement method and system for railway catenary based on vibration compensation. The measurement method involves constructing a category feature-filter parameter mapping lookup table by setting optimal filter parameters, performing a short-time Fourier transform on the online vibration signal, and extracting the online power spectral density as the online vibration feature vector. The method selects the operating condition corresponding to the cluster center with the smallest Euclidean distance as the current online operating condition and finds the corresponding optimal filter parameters. Kalman filtering vibration compensation is then performed using these optimal filter parameters as Kalman filtering parameters. The measurement system includes a signal acquisition module, a power spectral density extraction module, a dataset partitioning module, a mapping lookup table construction module, an online power spectral density extraction module, an optimal filter parameter lookup module, a Kalman filtering vibration compensation module, and an online adaptive update module. This invention can effectively filter and compensate vibration signals, making the final output result closer to the true value.

Description

Technical Field

[0001] This invention belongs to the technical field of railway maintenance equipment, and in particular relates to a method and system for measuring railway catenary based on vibration compensation. Background Technology

[0002] As a crucial component of railway electrification systems, the accuracy of the overhead contact system's geometric parameters (such as conductor height and pull-out value) directly impacts train operation safety. Currently, overhead contact system parameter measurement primarily employs laser-based overhead contact system measuring instruments, such as the subway overhead contact system geometric parameter measuring instrument disclosed in Chinese Patent CN202022932756.3. Workers push a measuring trolley equipped with a laser measuring component forward along the track to continuously measure the overhead contact system.

[0003] During operation, overhead contact line measuring devices are affected by multiple factors, including track irregularities, switch impacts, and vibrations from the vehicle's own mechanical structure. These complex vibration interferences can easily lead to distorted measurement results. Currently, vibration compensation technology for overhead contact line measuring devices mainly faces the following problems: Traditional compensation methods typically employ fixed-parameter Kalman filtering for vibration compensation, which cannot adapt to changes in vibration characteristics under different operating conditions. Based on practical experience with measuring devices, it has been found that the vibration characteristics of the measuring device are significantly correlated with environmental factors such as track gauge and single-section track length. When the measuring device operates in different measuring sections, its interference vibration spectrum characteristics change significantly. Fixed filter parameters lead to unstable compensation effects under different operating conditions, with compensation accuracy fluctuations reaching ±5.0 mm or more. Furthermore, existing compensation systems are usually optimized offline and cannot continuously optimize algorithm parameters based on actual measurement results. When track conditions change, the system cannot adaptively adjust, leading to a gradual deterioration in compensation effectiveness. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for measuring railway catenary based on vibration compensation. The technical solution adopted by this invention is as follows: The measurement method for railway catenary based on vibration compensation includes the following steps: S01. Operate the overhead contact line measuring device under typical line operating conditions to collect vibration signals and measured value signals; S02. Perform a short-time Fourier transform on the acquired raw signal and extract the power spectral density as the vibration feature vector. S03. Using the vibration feature vectors of different typical line operating conditions as a dataset, the K-means algorithm is used to divide the dataset into multiple categories. S04. Set a set of optimal filter parameters for each vibration category and construct a category feature-filter parameter mapping lookup table; S05. Perform online measurement of railway catenary, collect online vibration signals and online measurement value signals, perform short-time Fourier transform on the online vibration signals, and extract the online power spectral density as the online vibration feature vector; S06. Calculate the Euclidean distance between the online vibration feature vector and all cluster centers in the mapping lookup table. Select the working condition corresponding to the cluster center with the smallest Euclidean distance as the current online working condition. Find the optimal filter parameters corresponding to the current online working condition from the mapping lookup table. S07. Using the optimal filter parameters as the Kalman filter parameters, perform Kalman filter vibration compensation; S08. Calculate the energy of the compensated residual signal. If the residual energy continues to be higher than the preset energy threshold, perform online adaptive update and readjust the optimal filter parameters for this operating condition.

[0005] Preferably, S02 specifically includes the following steps: S021. Perform a windowed short-time Fourier transform on the vibration signal to convert the time-domain signal to the frequency domain. The overlap ratio is set to 75%, and a Hamming window function is used. The window function coefficients are... The expression is: ; Where N is the window length; The spectrum obtained from each STFT window is denoted as follows: , l =0,1,2…N-1 is the frequency domain index; S022. Divide typical sub-frequency bands, calculate the sub-frequency band power spectral density for the spectrum in each sub-frequency band, and merge them to form a vibration characteristic vector. The sub-frequency band power spectral density is the sum of the spectral energies of all spectra falling in that sub-frequency band.

[0006] Preferably, in S06, the online vibration feature vector f m The distance d between the i-th (i=1,2,…,8) cluster center i for: .

[0007] Preferably, the initial values ​​of the optimal filter parameters in S04 are calculated using a genetic algorithm, specifically including the following steps: S041. Set the parameter constraints for the genetic algorithm, initialize the population, and randomly generate M individuals; S042. Calculate the fitness value for each individual; S043. Sort the individuals according to their fitness, select the N individuals with the highest fitness, and perform crossover and mutation operations to form M new individuals. S044, Repeat S042-S043 until the maximum number of iterations is reached; S045. Select the individual with the highest fitness as the optimal parameter combination.

[0008] Preferably, the Kalman filter vibration compensation in S07 includes the following steps: S071. For the implementation of the overhead contact line measurement trolley, construct discrete state equations; S072. For the implementation of the overhead contact line measurement trolley, construct the measurement equation; S073. Perform prediction using Kalman filtering based on the current optimal parameters, the previous time step's posterior state estimate, and the posterior error covariance. S074. Calculate the innovation and the innovation covariance; S075. Calculate the Kalman gain; S076. Calculate the compensated measurement signal based on the Kalman gain correction state estimate.

[0009] Preferably, S08 specifically includes the following steps: S081. Calculate the energy of the compensated residual signal in real time within a sliding window, and compare the residual signal energy with the energy threshold. S082. If the residual signal energy exceeds the energy threshold for 5 consecutive seconds, then perform online adaptive update to update the optimal filter parameters corresponding to the current operating condition.

[0010] The preferred online adaptive update calculation process is as follows: a. Define the optimization objective; b. Calculate the descent gradient of the optimization objective using numerical differentiation; c. Numerically update the Kalman process noise covariance and the Kalman observation noise covariance.

[0011] A vibration-compensated railway catenary measurement system, used to implement the aforementioned vibration-compensated railway catenary measurement method, includes: The signal acquisition module is used to acquire vibration signals and measured value signals of the catenary measuring device operating under typical line conditions; The power spectral density extraction module is used to perform a short-time Fourier transform on the acquired raw signal and extract the power spectral density as a feature vector. The dataset partitioning module is used to divide the vibration feature vectors of different typical line operating conditions into multiple categories using the K-means algorithm; The mapping lookup table construction module is used to set a set of optimal filter parameters for each vibration category and construct a category feature-filter parameter mapping lookup table. The online power spectral density extraction module is used to collect online vibration signals and online measurement signals during the online measurement of railway catenary, perform short-time Fourier transform on the online vibration signals, and extract the online power spectral density as the online vibration feature vector; The optimal filter parameter lookup module is used to calculate the Euclidean distance between the online vibration feature vector and all cluster centers in the mapping lookup table, select the working condition corresponding to the cluster center with the smallest Euclidean distance as the current online working condition, and look up the optimal filter parameters corresponding to the current online working condition from the mapping lookup table. The Kalman filter vibration compensation module is used to perform Kalman filter vibration compensation using the optimal filter parameters as the Kalman filter parameters. The online adaptive update module is used to calculate the energy of the compensated residual signal. If the residual energy continues to be higher than the preset energy threshold, an online adaptive update is performed to readjust the optimal filter parameters for the current operating condition.

[0012] The beneficial effects of this invention are: This invention constructs a category feature-filter parameter mapping lookup table by setting optimal filter parameters. The mapping lookup table enables real-time and rapid lookup of the optimal filter parameters corresponding to the current online operating condition. Kalman filtering vibration compensation is performed using the optimal filter parameters as Kalman filtering parameters, which can effectively filter and compensate vibration signals with a large number of glitches, making the final output result closer to the true value. Online adaptive updating is achieved by adjusting the optimal filter parameters of the operating condition. Attached Figure Description

[0013] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the steps of the railway catenary measurement method based on vibration compensation according to Embodiment 1 of the present invention; Figure 2 This is a graph showing the measured values ​​before and after vibration compensation in Embodiment 1 of the present invention. Detailed Implementation

[0014] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0015] Example 1

[0016] Embodiment 1 of this invention provides a railway catenary measurement method based on vibration compensation. The overall concept is to classify the line conditions based on the differences in vibration distribution characteristics of different line sections, and then apply specific parameters for vibration compensation for different line conditions. The method includes the following steps: S01. Operate the overhead contact line measuring device under typical line operating conditions to collect vibration signals and measured value signals.

[0017] The typical track conditions refer to the fixed sections that the catenary measuring device usually measures. The vibration signals can be acquired and extracted using accelerometers or inertial measurement units (IMUs) on the measuring device. The measured values ​​are the catenary parameters measured by the vehicle's laser sensors. When acquiring vibration and measured values, as many different track sections and conditions as possible should be selected, such as various track gauges, single rail lengths, and whether the track passes through turnouts / bridge sections / curve sections, etc., so that subsequent condition classification is more comprehensive, rich, and targeted.

[0018] In this first embodiment, the sampling frequency is 1000Hz, the sampling duration is ≥5 minutes, and the obtained time-domain vibration signal is denoted as: ; Wherein, the sampling interval T s =0.001 seconds, k=0,1,2… is the time-domain sampling index.

[0019] In this embodiment, to simplify the calculation, the measured value signal is selected as the contact wire height measurement value. The same method can be used to compensate for the overhead contact line pull-out value.

[0020] S02. Perform a short-time Fourier transform (STFT) on the acquired raw signal to extract the power spectral density (PSD) as a feature vector. Specifically, this includes the following steps: S021. Perform a windowed short-time Fourier transform (STFT) on the vibration signal to convert the time-domain signal to the frequency domain.

[0021] The overlap ratio is set to 75%, and a Hamming window function is used. The window function coefficients are... The expression is: ; Where N is the window length, in this first embodiment, N=128, corresponding to a 128ms time window, and the frequency resolution is approximately 7.8Hz.

[0022] The spectrum obtained from each STFT window is denoted as follows: , l =0,1,2…N-1 is the frequency domain index.

[0023] The windowed short-time Fourier transform can be implemented directly using existing open-source code.

[0024] S022. Divide typical sub-frequency bands, calculate the sub-frequency band power spectral density (PSD) for the spectrum in each sub-frequency band, and form a vibration characteristic vector. The sub-frequency band power spectral density is the sum of the spectral energies of all spectra falling in that sub-frequency band.

[0025] In this first embodiment, the 1-1000Hz frequency band is divided into the following 6 sub-bands: 1-40Hz, 40-100Hz, 100-200 Hz, 200-400 Hz, 400-600 Hz, and 600-1000 Hz. The PSD value of vibration is calculated for each sub-band.

[0026] The energy of each spectrum is: ; The PSD value of the vibration in each sub-band is the sum of all spectral energies falling within that band. For example, for the first sub-band (1-40Hz), since the frequencies corresponding to X[0]~X[4] fall within the range of this sub-band, the PSD value of the first sub-band is: ; After calculating the PSD values ​​of the six sub-bands separately, they are combined and recorded as the vibration characteristic vector for this operating condition. f : .

[0027] The energy distribution of vibration is unique under different operating conditions (e.g., different track gauges, track lengths, etc.). This step characterizes different typical track operating conditions using vibration characteristic vectors, so that different compensation parameters can be used under different operating conditions.

[0028] S03, Vibration characteristic vectors for different typical line operating conditions f As a dataset, the K-means algorithm is used to divide the dataset into multiple categories.

[0029] The k-means algorithm is a widely used clustering algorithm used to divide data points into k clusters, such that points within the same cluster are similar to each other, while points in different clusters are as different as possible. The k-means algorithm can be implemented directly using open-source code.

[0030] In this embodiment, the number of categories k=8, meaning the k-means algorithm finds 8 centroids {c1,c2,...,c8} from the dataset, and each c i ∈R 6 , representing the center of the i-th cluster; the clustering iteration terminates when the maximum number of iterations reaches 100 or the sum of squared errors changes less than 1e-5; the clustering quality is evaluated using the silhouette coefficient.

[0031] This step involves analyzing a large number of vibration characteristic vectors from different typical line operating conditions. f The clusters are divided into 8 different categories, allowing for individual vibration compensation settings for each vibration category.

[0032] S04, Set a set of optimal filter parameters P for each vibration category. i * =(Q opt R opt ), where Q opt R is the optimal value of the noise covariance of the Kalman process. opt To observe the optimal value of noise covariance, a "category feature - filter parameter" mapping lookup table is constructed.

[0033] Kalman filtering is an efficient recursive estimation algorithm that uses state prediction, covariance prediction, and Kalman gain updates to iterate multiple times to make the optimal estimate of the true state of a dynamic system with noise interference.

[0034] The optimal filter parameter P i * The optimal values ​​for the Kalman filter parameters for each typical line operating condition can be determined manually by setting initial values ​​after multiple trials and then gradually optimizing and learning during actual use. Alternatively, an intelligent optimization algorithm can be used to set the initial values.

[0035] In this first embodiment, the optimal filter parameter P i * The initial value is calculated using a genetic algorithm, specifically including the following steps: S041. Set the genetic algorithm parameter constraint range, initialize the population, and randomly generate 50 individuals.

[0036] The "individual" refers to the initial filter parameter P. i =(Q, R), where the constraint range of the filter parameters is: Q∈[0.01, 10], R∈[0.1, 20]; S042. Calculate the fitness value for each individual. : ; in, The optimization objective, to compensate for the standard deviation of the measured residuals, is to maximize F. The calculation formula is: ; Let the innovation of the measurement values ​​before and after compensation at time k be denoted as , representing the deviation between the sensor's measured value and the predicted value: ; for The average value.

[0037] It directly reflects the stability of the filtered result. The smaller the value, the better the match between the predicted and measured values. The filter parameter P... i * The more reasonable the settings, the more significant the effect of vibration interference suppression.

[0038] S043. Sort individuals by fitness, select the 30 individuals with the highest fitness and perform crossover and mutation operations to form 50 new individuals; set the crossover probability to 0.8 and the mutation probability to 0.1.

[0039] S044. Repeat steps S042-S043 until the maximum number of iterations is reached; in this embodiment, the maximum number of iterations is 100.

[0040] S045. Select the individual with the highest fitness as the optimal parameter combination P. i * .

[0041] Genetic algorithms are mature intelligent optimization algorithms, and their implementation can utilize open-source code.

[0042] Finally, the optimal filter parameters P for each typical line operating condition are compared. i * A one-to-one correspondence is established and written into a "category feature - filter parameter" mapping lookup table. Typical line conditions after classification are clustered using cluster center c. i In this embodiment, eight types are divided into categories, and the mapping lookup table is represented as follows: ; The lookup table (LUT) can be stored in the FPGA's on-chip RAM for quick lookup during actual measurements. Typically, the lookup latency is <50μs, which meets the requirements for real-time comparison in practical applications.

[0043] The above steps are pre-training performed before the actual measurement.

[0044] S05. Perform online measurement of railway catenary, collect online vibration signals and online measurement value signals, perform short-time Fourier transform (STFT) on the online vibration signals, and extract the online power spectral density (PSD) as the online vibration feature vector.

[0045] The online measurement signal refers to the structure undergoing formal contact wire testing; the online vibration signal is denoted as... u m(k), the online measured value signal is denoted as y m (k).

[0046] The specific calculation process of the short-time Fourier transform is exactly the same as step S02 (only the input variables are different), and the calculation result is the online vibration feature vector. f m In this first embodiment, f m The PSD values ​​of the six sub-bands of the online vibration signal.

[0047] In this embodiment, the STFT window length is 128. Existing hardware devices can perform short-time Fourier transforms with a computation delay of less than 50μs, which is sufficient to meet the requirement of real-time selection of the optimal filter parameter P. i * Requirements.

[0048] S06. Calculate the online vibration characteristic vector. f m The operating condition corresponding to the cluster center with the smallest Euclidean distance from all cluster centers in the mapping lookup table is selected as the current online operating condition. The optimal filter parameter P corresponding to the current online operating condition is then found from the mapping lookup table. i * .

[0049] In this first embodiment, the online vibration feature vector f m The distance d between the i-th (i=1,2,…,8) cluster center i for: ; Through the above steps, this method can classify operating conditions in real time based on online vibration signals and find the optimal filter parameter P by looking up a table. i * This allows for the selection of the most suitable filter parameters for any given measurement time, achieving efficient and reliable filtering.

[0050] S07, with optimal filter parameters P i * Use the Kalman filter parameters to perform Kalman filter vibration compensation.

[0051] Kalman filter vibration compensation includes the following steps: S071. For the implementation of the overhead contact line measurement trolley, the discrete state equation is constructed as follows: ; in, Let F be a second-order state vector containing the actual height of the overhead contact line and its rate of change, where F is the state transition matrix and B is the control matrix of the state. ; For process noise, The covariance of the process noise. For the actual height of the overhead contact line, This represents the rate of change of the overhead contact line height.

[0052] S072. For the implementation of the overhead contact line measurement trolley, construct the measurement equation: ; Where H is the observation matrix, . R represents the observation noise, and R represents the observation noise covariance.

[0053] S073, using the current optimal parameters Q and R and the posterior state estimate from the previous time step. Posterior error covariance Perform prediction using basic Kalman filtering: ; ; in For prior state estimation, Let be the prior error covariance.

[0054] S074, Calculating New Information and new information covariance : ; ; Among them, new information Information covariance represents the deviation between the measured and predicted values ​​of a sensor. Used to depict new information The magnitude of uncertainty.

[0055] S075. Calculate Kalman gain : ; S076, Based on Kalman gain To correct the state estimate, calculate the compensated overhead contact line conductor height value. : ; ; in, For posterior state estimation, This is the posterior error covariance, used for iteration at the next time step. The compensated overhead contact line height output is: ; Compensation results This is the conductor height of the contact network after filtering.

[0056] Using the above method, vibration compensation can be performed on the contact wire measurement signal through Kalman filtering, resulting in a contact wire conduction height output that is closer to the true value.

[0057] The above algorithm can be implemented using existing algorithm frameworks, such as calling the filterpy library and setting parameters according to the above formula.

[0058] like Figure 2 As shown in the figure, the simulated signal and the Kalman vibration-compensated signal obtained through Matlab measurement are illustrated. The blue curve represents the measured contact wire height signal, and the red curve represents the contact wire height curve after vibration compensation. The comparison demonstrates that this invention, through optimized Kalman filtering parameters, can effectively filter and compensate for vibration signals with numerous spikes, resulting in a smoother final output that more closely approximates the true value.

[0059] S08. Calculate the energy of the compensated residual signal. If the residual energy continues to be higher than the preset energy threshold, perform online adaptive update and readjust the optimal filter parameter P for this operating condition. i * Specifically, it includes the following steps: S081, in a space of length N E Real-time calculation of the compensated residual signal energy within the sliding window The residual signal energy is then compared with the energy threshold.

[0060] The residual signal energy of the k-th sliding window The definition is as follows: ; In this embodiment, N is the new information. E Take 500 (corresponding to a 0.5-second time window).

[0061] S082, If the residual signal energy Exceeding the energy threshold for 5 consecutive seconds Then, online adaptive updates are performed to adjust the optimal filter parameters P corresponding to the current operating condition. i * Update.

[0062] Energy threshold The selection can be based on historical experience data or offline data statistics.

[0063] The calculation process for online adaptive updates is as follows: a. Define the optimization objective: ; The optimization objective is to adjust Q and R to reduce the residual signal energy. minimize.

[0064] b. Using numerical differentiation calculation The descent gradient: ; ; in, and These represent tiny increases or decreases in Q and R (e.g., 0.01). The calculation here uses a trial-and-error method to determine the effect of small changes in Q and R. The change in makes Decrease.

[0065] c. Numerically update the Kalman process noise covariance Q and the Kalman observation noise covariance R: ; ; in The learning rate is selected based on project requirements. A higher learning rate results in faster update speeds, but also lower stability and learning rate. Too high a value can cause fluctuations in the update results, so a value of 0.005 to 0.1 is generally chosen.

[0066] Residual signal energy If the residual signal energy remains too high for an extended period, it indicates a decrease in the compensation effect (e.g., a change in operating conditions). Therefore, based on the gradient descent concept, this invention further designs an online adaptive update mechanism. When the residual signal energy remains too high, online learning is performed through the above steps to update the parameters Q and R, ensuring a good compensation effect.

[0067] Example 2

[0068] Embodiment 2 of the present invention provides a railway catenary measurement system based on vibration compensation, used to implement the railway catenary measurement method based on vibration compensation described in Embodiment 1, including: The signal acquisition module is used to implement step S01 of Embodiment 1; The power spectral density extraction module is used to implement step S02 of Example 1; The dataset partitioning module is used to implement step S03 of Example 1; The mapping lookup table construction module is used to implement step S04 of embodiment one; An online power spectral density extraction module is used to implement step S05 of Example 1; The optimal filter parameter lookup module is used to implement step S06 of Embodiment 1. The Kalman filter vibration compensation module is used to implement step S07 of Embodiment 1. The online adaptive update module is used to implement step S08 of Embodiment 1.

[0069] In the embodiments of the present invention, all technical features not described in detail are existing technologies or conventional technical means, and will not be repeated here.

[0070] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A method for measuring railway catenary based on vibration compensation, characterized in that, Includes the following steps: S01. Operate the overhead contact line measuring device under typical line operating conditions to collect vibration signals and measured value signals; S02. Perform a short-time Fourier transform on the acquired raw signal and extract the power spectral density as the vibration feature vector. S03. Using the vibration feature vectors of different typical line operating conditions as a dataset, the K-means algorithm is used to divide the dataset into multiple categories. S04. Set a set of optimal filter parameters for each vibration category and construct a category feature-filter parameter mapping lookup table; S05. Perform online measurement of railway catenary, collect online vibration signals and online measurement value signals, perform short-time Fourier transform on the online vibration signals, and extract the online power spectral density as the online vibration feature vector; S06. Calculate the Euclidean distance between the online vibration feature vector and all cluster centers in the mapping lookup table. Select the working condition corresponding to the cluster center with the smallest Euclidean distance as the current online working condition. Find the optimal filter parameters corresponding to the current online working condition from the mapping lookup table. S07. Using the optimal filter parameters as the Kalman filter parameters, perform Kalman filter vibration compensation; S08. Calculate the energy of the compensated residual signal. If the residual energy continues to be higher than the preset energy threshold, perform online adaptive update and readjust the optimal filter parameters for this operating condition.

2. The railway catenary measurement method based on vibration compensation according to claim 1, characterized in that, S02 specifically includes the following steps: S021. Perform a windowed short-time Fourier transform on the vibration signal to convert the time-domain signal to the frequency domain. The overlap ratio is set to 75%, and a Hamming window function is used. The window function coefficients are... The expression is: ； Where N is the window length; The spectrum obtained from each STFT window is denoted as follows: , l =0,1,2…N-1 is the frequency domain index; S022. Divide typical sub-frequency bands, calculate the sub-frequency band power spectral density for the spectrum in each sub-frequency band, and merge them to form a vibration characteristic vector. The sub-frequency band power spectral density is the sum of the spectral energies of all spectra falling in that sub-frequency band.

3. The railway catenary measurement method based on vibration compensation according to claim 1, characterized in that, In S06, the online vibration feature vector f m The distance d between the i-th (i=1,2,…,8) cluster center i for: 。 4. The railway catenary measurement method based on vibration compensation according to claim 1, characterized in that, The initial values ​​of the optimal filter parameters in S04 are calculated using a genetic algorithm, specifically including the following steps: S041. Set the parameter constraints for the genetic algorithm, initialize the population, and randomly generate M individuals; S042. Calculate the fitness value for each individual; S043. Sort the individuals according to their fitness, select the N individuals with the highest fitness, and perform crossover and mutation operations to form M new individuals. S044, Repeat S042-S043 until the maximum number of iterations is reached; S045. Select the individual with the highest fitness as the optimal parameter combination.

5. The railway catenary measurement method based on vibration compensation according to claim 1, characterized in that, Kalman filter vibration compensation in S07 includes the following steps: S071. For the implementation of the overhead contact line measurement trolley, construct discrete state equations; S072. For the implementation of the overhead contact line measurement trolley, construct the measurement equation; S073. Perform prediction using Kalman filtering based on the current optimal parameters, the previous time step's posterior state estimate, and the posterior error covariance. S074. Calculate the innovation and the innovation covariance; S075. Calculate the Kalman gain; S076. Calculate the compensated measurement signal based on the Kalman gain correction state estimate.

6. The railway catenary measurement method based on vibration compensation according to claim 1, characterized in that, S08 specifically includes the following steps: S081. Calculate the energy of the compensated residual signal in real time within a sliding window, and compare the residual signal energy with the energy threshold. S082. If the residual signal energy exceeds the energy threshold for 5 consecutive seconds, then perform online adaptive update to update the optimal filter parameters corresponding to the current operating condition.

7. The railway catenary measurement method based on vibration compensation according to claim 6, characterized in that, The calculation process for online adaptive updates is as follows: a. Define the optimization objective; b. Calculate the descent gradient of the optimization objective using numerical differentiation; c. Numerically update the Kalman process noise covariance and the Kalman observation noise covariance.

8. A vibration-compensated railway catenary measurement system, used to implement the vibration-compensated railway catenary measurement method as described in claim 1, comprising: The signal acquisition module is used to acquire vibration signals and measured value signals of the catenary measuring device operating under typical line conditions; The power spectral density extraction module is used to perform a short-time Fourier transform on the acquired raw signal and extract the power spectral density as a feature vector. The dataset partitioning module is used to divide the vibration feature vectors of different typical line operating conditions into multiple categories using the K-means algorithm; The mapping lookup table construction module is used to set a set of optimal filter parameters for each vibration category and construct a category feature-filter parameter mapping lookup table. The online power spectral density extraction module is used to collect online vibration signals and online measurement signals during the online measurement of railway catenary, perform short-time Fourier transform on the online vibration signals, and extract the online power spectral density as the online vibration feature vector; The optimal filter parameter lookup module is used to calculate the Euclidean distance between the online vibration feature vector and all cluster centers in the mapping lookup table, select the working condition corresponding to the cluster center with the smallest Euclidean distance as the current online working condition, and look up the optimal filter parameters corresponding to the current online working condition from the mapping lookup table. The Kalman filter vibration compensation module is used to perform Kalman filter vibration compensation using the optimal filter parameters as the Kalman filter parameters. The online adaptive update module is used to calculate the energy of the compensated residual signal. If the residual energy continues to be higher than the preset energy threshold, an online adaptive update is performed to readjust the optimal filter parameters for the current operating condition.

Images