A method and system for associating positioning information based on multi-source data with a train

Abstract

The present invention discloses a method and system for associating positioning information with a train based on multi-source data, including collecting train operation data from multiple data sources and preprocessing the train operation data; performing train number clustering on the train operation data to obtain train movement data, and performing train number stringing on the train movement data to obtain the train running route; performing positioning integrity monitoring on the train running route to obtain integrity data, and correcting the integrity data according to the real-time data to obtain monitored and corrected data; calculating the train positioning according to the monitored and corrected data, and constructing an association model between the positioning information and the train according to the train positioning; optimizing the association model between the positioning information and the train according to the positioning error, inputting the data to be associated into the association model between the positioning information and the train, and outputting an association result. This method can improve the accuracy of associating the positioning information with the train, and at the same time has good interpretability, and can be directly applied to the train information positioning system.

Description

Technical Field

[0001] The present invention relates to the field of multi-source positioning, and particularly to a method and system for associating positioning information based on multi-source data with trains. Background Art

[0002] With the rapid development of the railway transportation industry, the requirements for the safety and real-time performance of the official document transportation process are increasing day by day. Traditional official document tracking methods often rely on manual records and regular inspections, which are not only inefficient but also difficult to meet the needs of modern transportation management. Although significant progress has been made in GPS positioning technology in the field of logistics tracking in recent years, in the train transportation scenario, when the train is running in tunnels or blocked by urban high-rise buildings, the GPS signal is vulnerable to interference, resulting in a decrease in positioning accuracy or even a complete loss of signal in some areas.

[0003] To make up for the deficiencies of GPS positioning, some position correction methods based on other auxiliary technologies have emerged in the market, such as using data from train operation control systems, track circuit information, or train wireless communication systems to assist in positioning. However, these methods still have room for improvement in terms of real-time performance, accuracy, and system integration. The present invention integrates multiple data sources such as GPS positioning information, WTDS data, train timetables, and operation line data to construct a set of highly efficient collaborative solutions, aiming to improve the real-time performance and accuracy of train positioning and official document tracking. This integrated strategy not only strengthens the security defense line of official documents but also realizes comprehensive traceability from the source to the terminal, setting a new benchmark for the seamless flow and security guarantee of information. Summary of the Invention

[0004] The object of the present invention is to provide a method for associating positioning information based on multi-source data with trains.

[0005] To achieve the above object, the present invention is implemented according to the following technical solution:

[0006] The present invention includes the following steps:

[0007] Collect train operation data from multiple data sources and preprocess the train operation data; the train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data;

[0008] Perform train number clustering on the train operation data to obtain train movement data, and perform train number stringing on the train movement data to obtain train running routes;

[0009] Perform positioning integrity monitoring on the train running routes to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data;

[0010] Calculate the train positioning based on the monitored correction data, and construct a positioning information and train association model based on the train positioning; the train positioning is obtained by using the global navigation satellite system through the signal propagation time;

[0011] Optimize the positioning information and train association model according to the positioning error, input the data to be associated into the positioning information and train association model, and output the association result.

[0012] Furthermore, the preprocessing method includes:

[0013] Process and group the train operation data according to each dispatching section; compare the train operation data of the dispatching section with the section dictionary and correct the section data; generate the description information corresponding to the section according to the section data, and the description information includes: train number ID, dispatching desk ID, running train number, starting station, starting time, ending station, ending time.

[0014] Furthermore, the method for obtaining the train running data by clustering the train numbers of the train operation data includes:

[0015] Perform conditional judgment on the preprocessed train operation data for the corresponding train number class, and the judgment conditions include checking whether the running train number is included, whether the running stations are close, and whether the time span exceeds a preset threshold;

[0016] If the conditional judgment of the corresponding train number class is satisfied, classify the full-section data of the train operation data and update the cluster information;

[0017] If the conditional judgment of the corresponding train number class is not satisfied, generate a new cluster and class description information, and the class description information includes: class ID, running train number set; where the running train number set includes starting station, starting time, ending station, ending time, dispatching desk, train number ID;

[0018] Traverse all the train operation data, and output the new cluster and class description information as the train running data.

[0019] Furthermore, the method for obtaining the train running path by stringing the train numbers of the train running data includes:

[0020] String the train running data in the same cluster through the coherence of time and space;

[0021] Customize the splitting rules according to the properties of different types of trains, and split the stringing result through the splitting rules to obtain the train running path of a single operation.

[0022] Furthermore, the method for obtaining the integrity data by monitoring the positioning integrity of the train running path includes:

[0023] Train positioning is obtained according to the GPS global satellite positioning combined with the track potential positioning method to obtain the train running route;

[0024] A filtering parameter selection method based on a fuzzy inference system is adopted. The fuzzy inference system dynamically adjusts the filtering parameters according to the characteristics of the train positioning data to realize the integrity monitoring of the train running route.

[0025] Furthermore, the method for correcting the integrity data according to the real-time data to obtain the monitoring correction data includes:

[0026] Calculate the similarity between the real-time data and the integrity data:

[0027]

[0028] Where the i-th real-time data is a i , the i-th integrity data is f i , and the similarity between the real-time data a and the integrity data f is The influence factor is ζ, the regulation constant is x, and the number of real-time data is The number of integrity data is

[0029] Pair the integrity data and the real-time data with a similarity greater than 0.713, and construct a recurrence relationship about the time series according to the real-time data. The expression is:

[0030] a i (s + 1) = log2(e -ζ + 0.12)·a i T (s) + σ 2 (s + 1)

[0031] Where the i-th real-time data at the s + 1-th moment is The error between the integrity data and the real-time data at the s + 1-th moment is σ(s + 1), and the i-th real-time data at the s-th moment is a i (s), transposed to T;

[0032] Use the least squares algorithm for recurrence. The expression is:

[0033]

[0034] Where the gain matrix at the s-th moment is The covariance matrix at the s-th moment is The suppression factor is ρ, and the covariance matrix at the s + 1-th moment is

[0035] Update the real-time data. The expression is:

[0036]

[0037] Among them, the updated i-th real-time data at the (s + 1)-th moment is a i (s + 1), and the penalty weight at the s-th moment is

[0038] Calculate the error ratio parameter between the real-time data and the integrity data:

[0039]

[0040] Among them, the error ratio parameter between the i-th real-time data and the integrity data at the (s + 1)-th moment is K(s + 1);

[0041] Correct the integrity data according to the error ratio parameter:

[0042]

[0043] Among them, the error correction factor is ξ, the optimal control coefficient is ξ, and the i-th actual integrity data value is f i , and the corrected i-th integrity data is

[0044] Furthermore, the method for constructing the positioning information and train association model according to the train positioning includes:

[0045] Calculate the loss function of the actual train positioning and the predicted train positioning, and take the minimum value of the loss function as the objective function of the positioning information and train association model;

[0046] The positioning information and train association model includes a clustering anomaly recognition algorithm, a causal inference method, and a deep belief network algorithm;

[0047] The clustering anomaly recognition algorithm groups the input multi-source train operation data into different clusters, identifies the data points in the multi-source train operation data that are significantly different from the main cluster, and marks the significantly different data as abnormal data;

[0048] The causal inference method determines the cause and consequence of the anomaly by analyzing the causal relationship between the abnormal data and the train positioning variables, quantifies the impact of the anomaly on the system or data set according to the cause and consequence, and obtains the impact coefficient;

[0049] The deep belief network algorithm performs layer-by-layer pre-training and fine-tuning of the deep neural network through the multi-source train operation data and the impact coefficient, learns complex non-linear relationships, and predicts the train positioning information.

[0050] Furthermore, the method for optimizing the positioning information and train association model according to the positioning error includes:

[0051] Introduce a particle population, use the positioning error as the fitness function, calculate the fitness of the particle population, and take the particle with the minimum fitness as the optimal solution;

[0052] Take the minimum positioning error as the search strategy, and initialize the particle population. The expression is:

[0053]

[0054] where the position of the z-th particle in the d-th dimension is The random number from 0 to 1 is τ1, and the lower limit of the d-th dimension is The upper limit of the d-th dimension is

[0055] Update the particle position. The expression is:

[0056]

[0057] where the position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is The optimal solution is The random integer from 1 to 2 is γ, and the fitness function value of the optimal solution is The fitness function value of the z-th particle is The current iteration number is t, and the maximum iteration number is t max , and the random numbers from 0 to 1 are τ2 and τ3 respectively;

[0058] Update the particle position to obtain the search position. The expression is:

[0059]

[0060] where the search position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is The position of the z-th particle in the d-th dimension at the t-th iteration is The random number from 0 to 1 is τ4;

[0061] Use the attenuation factor to update the search position to obtain the adaptation position. The expression is:

[0062]

[0063] where the search position of the z-th particle in the d-th dimension at the t-th iteration is The random number from 0 to 1 is τ5, the attenuation factor is ε, and the adaptation position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is

[0064] Use non-linear adaptive adjustment of the attenuation factor. The expression is:

[0065]

[0066] Among them, the maximum value of the attenuation factor is ε max , and the minimum value of the attenuation factor is ε min , and the attenuation factor for the t-th iteration is ε(t + 1);

[0067] Update the adaptation position using the updated attenuation factor, and continuously iterate until the positioning error reaches the minimum. Otherwise, update the particles and readjust the attenuation factor.

[0068] In a second aspect, a system for associating positioning information based on multi-source data with a train includes:

[0069] Data acquisition module: used to acquire train operation data from multiple data sources and preprocess the train operation data; the train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data;

[0070] Clustering and stringing module: used to perform train number clustering on the train operation data to obtain train walking data, and perform train number stringing on the train walking data to obtain the train running route;

[0071] Monitoring and correction module: used to monitor the positioning integrity of the train running route to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data;

[0072] Calculation and construction module: used to calculate the train positioning according to the monitored and corrected data, and construct an association model between positioning information and the train based on the train positioning;

[0073] Optimization and output module: used to optimize the association model between positioning information and the train according to the positioning error, input the data to be associated into the association model between positioning information and the train, and output the association result.

[0074] The beneficial effects of the present invention are:

[0075] The present invention is a method and system for associating positioning information based on multi-source data with a train. Compared with the prior art, the present invention has the following technical effects:

[0076] Through the steps of preprocessing, train trip clustering, train trip stringing, positioning integrity monitoring, data correction, model construction, and model optimization, the present invention can improve the accuracy of associating positioning information based on multi-source data with trains, thereby enhancing the precision of associating positioning information based on multi-source data with trains. Optimizing the association of positioning information based on multi-source data with trains can greatly save resources and improve work efficiency. It can achieve intelligent association of positioning information based on multi-source data with trains, perform real-time train trip clustering stringing and data correction on the association of positioning information based on multi-source data with trains, which is of great significance for the association of positioning information based on multi-source data with trains. It can adapt to the association of positioning information based on multi-source data with trains of different standards and the association requirements of different multi-source data with trains, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a flowchart of the steps of a method for associating positioning information based on multi-source data with trains according to the present invention;

[0078] Figure 2 It is a flowchart of train operation line data processing according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] The present invention will be further described below through specific embodiments. The illustrative embodiments and explanations of the present invention are used to explain the present invention, but do not limit the present invention.

[0080] A method and system for associating positioning information based on multi-source data with trains according to the present invention include the following steps:

[0081] As Figure 1 shown, in this embodiment, it includes the following steps:

[0082] Collect train operation data from multiple data sources and preprocess the train operation data; the train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data;

[0083] In actual evaluation, train A and train B at XXX Railway Station are used as research objects; multiple data sources include GPS, WTDS, train timetables, and operation line data;

[0084] Historical data includes train route information, train timetables, train operation routes, train number IDs, dispatching desk IDs, running train numbers, starting stations, starting times, ending stations, ending times, train speeds, and train positions;

[0085] Real-time data includes train number IDs, dispatching desk IDs, running train numbers, starting stations, starting times, ending stations, ending times, current train speeds, and current train positions;

[0086] Perform train number clustering on the train operation data to obtain train movement data, and perform train number stringing on the train movement data to obtain the train running route;

[0087] In the actual evaluation, the train running route: Train A: Station 20 > Station 21 >... > Station 29 > Station 30 >... > Station 20;

[0088] Train B: Station 25 > Station 26 >... > Station 34 > Station 35 >... > Station 25;

[0089] Perform positioning integrity monitoring on the train running route to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data;

[0090] In the actual evaluation, the positioning integrity of Train A is 0.982, and the positioning integrity of Train B is 0.973;

[0091] Calculate the train positioning according to the monitored and corrected data, and construct a positioning information and train association model according to the train positioning; the train positioning is obtained by using the global navigation satellite system through the signal propagation time;

[0092] In the actual evaluation, due to the construction ahead, the actual acceleration of Train A is reduced to 0.3 m / s 2 , and the corrected integrity is 95%; the position data of Train B is normal and does not need to be corrected;

[0093] Optimize the positioning information and train association model according to the positioning error, input the data to be associated into the positioning information and train association model, and output the association result.

[0094] In this embodiment, the preprocessing method includes:

[0095] Process and group the train operation data according to each dispatching station section; compare the train operation data of the dispatching section with the section dictionary, and correct the section data; generate the description information corresponding to the section according to the section data, and the description information includes: train number ID, dispatching station ID, running train number, starting station, starting time, ending station, ending time.

[0096] In this embodiment, the method for performing train number clustering on the train operation data to obtain train movement data includes:

[0097] Perform conditional judgment on the preprocessed train operation data for the corresponding train number class, and the judgment conditions include checking whether the running train number is included, whether the running stations are close, and whether the time span exceeds a preset threshold;

[0098] If the condition judgment for the corresponding train class is satisfied, then classify all the train operation data and update the cluster information;

[0099] If the condition judgment for the corresponding train class is not satisfied, then generate new cluster and class description information. The class description information includes: class ID, set of running train numbers; where the set of running train numbers includes starting station, starting time, ending station, ending time, dispatching desk, train number ID;

[0100] Traverse all the train operation data and output the new cluster and class description information as train movement data.

[0101] In this embodiment, the method for obtaining the train running path by stringing train numbers for the train movement data includes:

[0102] String together the train movement data in the same cluster through the coherence of time and space;

[0103] Customize splitting rules according to the nature of different types of trains, and split the stringing result through the splitting rules to obtain the train running path of a single operation.

[0104] In this embodiment, the method for obtaining integrity data by monitoring the integrity of the train running path includes:

[0105] Obtain the train positioning of the train running path according to GPS global satellite positioning combined with the track potential positioning method;

[0106] Adopt a filtering parameter selection method based on a fuzzy inference system, and use the fuzzy inference system to dynamically adjust the filtering parameters according to the characteristics of the train positioning data to achieve the integrity monitoring of the train running path.

[0107] In this embodiment, the method for obtaining monitoring correction data by correcting the integrity data according to the real-time data includes:

[0108] Calculate the similarity between the real-time data and the integrity data:

[0109]

[0110] Where the i-th real-time data is a i , the i-th integrity data is f i , and the similarity between the real-time data a and the integrity data f is The influence factor is ζ, the regulation constant is x, the number of real-time data is The number of integrity data is

[0111] Pair the integrity data with a similarity greater than 0.713 with the real-time data, and construct a recurrence relation for the time series based on the real-time data. The expression is:

[0112]

[0113] Among them, the i-th real-time data at the s + 1-th moment is The error between the integrity data and the real-time data at the s + 1-th moment is σ(s + 1), and the i-th real-time data at the s-th moment is a i (s), and the transpose is T;

[0114] Use the least squares algorithm for recurrence. The expression is:

[0115]

[0116] Among them, the gain matrix at the s-th moment is The covariance matrix at the s-th moment is The suppression factor is ρ, and the covariance matrix at the s + 1-th moment is

[0117] Update the real-time data. The expression is:

[0118]

[0119] Among them, the updated i-th real-time data at the s + 1-th moment is á i (s + 1), and the penalty weight at the s-th moment is

[0120] Calculate the error ratio parameter between the real-time data and the integrity data:

[0121]

[0122] Among them, the error ratio parameter between the i-th real-time data and the integrity data at the s + 1-th moment is K(s + 1);

[0123] Correct the integrity data according to the error ratio parameter:

[0124]

[0125] Among them, the error correction factor is ξ, the optimal control coefficient is ξ, and the i-th actual integrity data value is The corrected i-th integrity data is

[0126] In this embodiment, the method for constructing the positioning information and train association model according to the train positioning includes:

[0127] Calculate the loss function of the actual train positioning and the predicted train positioning, and take the minimum value of the loss function as the objective function of the positioning information and the train association model;

[0128] The positioning information and train association model includes a clustering anomaly recognition algorithm, a causal inference method, and a deep belief network algorithm;

[0129] The clustering anomaly recognition algorithm groups the input multi-source train operation data into different clusters, identifies the data points in the multi-source train operation data that are significantly different from the main cluster, and marks the significantly different data as abnormal data;

[0130] The causal inference method determines the causes and consequences of anomalies by analyzing the causal relationship between abnormal data and train positioning variables, quantifies the impact of anomalies on the system or dataset based on the causes and consequences, and obtains the impact coefficient;

[0131] The deep belief network algorithm performs layer-by-layer pre-training and fine-tuning of the deep neural network through multi-source train operation data and the impact coefficient, learns complex non-linear relationships, and predicts train positioning information.

[0132] In this embodiment, the method for optimizing the positioning information and train association model according to the positioning error includes:

[0133] Introduce a particle swarm, use the positioning error as the fitness function, calculate the fitness of the particle swarm, and take the particle with the minimum fitness as the optimal solution;

[0134] Take the minimum positioning error as the search strategy to initialize the particle swarm. The expression is:

[0135]

[0136] where the position of the z-th particle in the d-th dimension is The random number from 0 to 1 is τ1, and the lower limit of the d-th dimension is The upper limit of the d-th dimension is

[0137] Update the particle position. The expression is:

[0138]

[0139] where the position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is The optimal solution is The random integer from 1 to 2 is γ, and the fitness function value of the optimal solution is The fitness function value of the z-th particle is The current iteration number is t, and the maximum iteration number is t max , and the random numbers from 0 to 1 are τ2 and τ3 respectively;

[0140] Update the particle position to obtain the search position, and the expression is:

[0141]

[0142] where the search position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is The position of the z-th particle in the d-th dimension at the t-th iteration is The random number from 0 to 1 is τ4;

[0143] Update the search position using the attenuation factor to obtain the adaptation position, and the expression is:

[0144]

[0145] where the search position of the z-th particle in the d-th dimension at the t-th iteration is The random number from 0 to 1 is τ5, the attenuation factor is ε, and the adaptation position of the z-th particle in the d-th dimension at the (t + 1)-th iteration is

[0146] Adopt a non-linear adaptive adjustment of the attenuation factor, and the expression is:

[0147]

[0148] where the maximum value of the attenuation factor is ε max , the minimum value of the attenuation factor is ε min , and the attenuation factor at the t-th iteration is ε(t + 1);

[0149] Update the adaptation position using the updated attenuation factor, and iterate continuously until the positioning error reaches the minimum, otherwise update the particles and readjust the attenuation factor.

[0150] In a second aspect, a system for associating positioning information based on multi-source data with a train includes:

[0151] Data acquisition module: used to acquire the train operation data of multiple data sources and preprocess the train operation data; the train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data;

[0152] Clustering and threading module: used to perform train number clustering on the train operation data to obtain train walking data, and perform train number threading on the train walking data to obtain the train running route;

[0153] Monitoring and correction module: used to monitor the integrity of the train running route to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data;

[0154] Calculation and construction module: used to calculate the train positioning according to the monitored and corrected data, and construct a positioning information and train association model according to the train positioning;

[0155] Optimization and output module: used to optimize the positioning information and train association model according to the positioning error, input the data to be associated into the positioning information and train association model, and output the association result.

[0156] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for associating positioning information based on multi-source data with a train, characterized in that, Including the following steps: Collect train operation data from multiple data sources and preprocess the train operation data; the train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data; Perform train number clustering on the train operation data to obtain train walking data, and perform train number stringing on the train walking data to obtain train running routes; Perform positioning integrity monitoring on the train running routes to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data; Calculate train positioning according to the monitored and corrected data, and construct a positioning information and train association model according to the train positioning; the train positioning is obtained by using the global navigation satellite system through signal propagation time; Optimize the positioning information and train association model according to the positioning error, input the data to be associated into the positioning information and train association model, and output an association result; including: Calculate the loss function of the actual train positioning and the predicted train positioning, and use the minimum value of the loss function as the objective function of the positioning information and train association model; The positioning information and train association model includes a clustering anomaly recognition algorithm, a causal inference method, and a deep belief network algorithm; The clustering anomaly recognition algorithm groups the input multi-source train operation data into different clusters, identifies data points in the multi-source train operation data that are significantly different from the main cluster, and marks the significantly different data as abnormal data; The causal inference method determines the causes and consequences of anomalies by analyzing the causal relationship between abnormal data and train positioning variables, quantifies the impact of anomalies on the system or dataset according to the causes and consequences, and obtains an impact coefficient; The deep belief network algorithm performs layer-by-layer pre-training and fine-tuning of the deep neural network through multi-source train operation data and the impact coefficient, learns complex non-linear relationships, and predicts train positioning information.

2. The method for associating positioning information based on multi-source data with a train according to claim 1, characterized in that The method of the preprocessing includes: Process and group the train operation data according to each dispatching station section; compare the train operation data of the dispatching section with the section dictionary and correct the section data; generate description information corresponding to the section according to the section data, and the description information includes: train number ID, dispatching station ID, running train number, starting station, starting time, ending station, ending time.

3. The method for associating positioning information based on multi-source data with a train according to claim 1, wherein The method of performing train number clustering on the train operation data to obtain train walking data includes: Perform conditional judgment on the preprocessed train operation data for the corresponding train number class, and the judgment conditions include checking whether the running train number is included, whether the running stations are close, and whether the time span exceeds a preset threshold; If the conditional judgment of the corresponding train number class is satisfied, classify all the data of the train operation data and update the cluster information; If the conditional judgment of the corresponding train number class is not satisfied, generate a new cluster and class description information, and the class description information includes: class ID, running train number set; where the running train number set includes starting station, starting time, ending station, ending time, dispatching station, train number ID; Traverse all the train operation data, and output the new cluster and class description information as train walking data.

4. The method for associating positioning information based on multi-source data with a train according to claim 1, wherein A method for obtaining the train running path by stringing train numbers for the train running data, including: Stringing the train running data in the same cluster through the coherence of time and space; Customizing a splitting rule according to the nature of different types of trains, and splitting the stringing result through the splitting rule to obtain the train running path of a single operation.

5. The method for associating positioning information based on multi-source data with a train according to claim 1, wherein A method for obtaining integrity data by monitoring the positioning integrity of the train running path, including: Obtaining the train positioning of the train running path according to GPS global satellite positioning combined with the track potential positioning method; Adopting a filtering parameter selection method based on a fuzzy inference system, and using the fuzzy inference system to dynamically adjust the filtering parameters according to the characteristics of the train positioning data to achieve the integrity monitoring of the train running path.

6. The method for associating positioning information based on multi-source data with a train according to claim 1, wherein A method for correcting the integrity data according to the real-time data to obtain monitored and corrected data, including: Calculating the similarity between the real-time data and the integrity data: ； where the i-th real-time data is , the i-th integrity data is , the similarity between the real-time data a and the integrity data f is , the influencing factor is , the regulation constant is x, the number of real-time data is , the number of integrity data is ; Pairing the integrity data and the real-time data with a similarity greater than 0.713, and constructing a recurrence relationship about the time series according to the real-time data, and the expression is: ； Among them, the i-th real-time data at the +1 moment is , and the error between the integrity data and the real-time data at the +1 moment is . The i-th real-time data at the moment is , and the transpose is T; Using the least squares algorithm for recurrence, and the expression is: ； where the gain matrix at the s-th moment is , the covariance matrix at the s-th moment is , the suppression factor is , and the covariance matrix at the (s + 1)-th moment is ; Updating the real-time data, and the expression is: ； where the updated $i$-th real-time data at the $(s + 1)$-th moment is , and the penalty weight at the $s$-th moment is ; Calculating the error ratio parameter between the real-time data and the integrity data: ； where the error ratio parameter between the i-th real-time data and the integrity data at the (s + 1)-th moment is ; Correcting the integrity data according to the error ratio parameter: ； where the error correction factor is , the i-th actual integrity data value is , and the i-th integrity data after correction is .

7. The method for associating positioning information based on multi-source data with a train according to claim 1, wherein A method for optimizing the association model between the positioning information and the train according to the positioning error, including: Introducing a particle population, taking the positioning error as the fitness function, calculating the fitness of the particle population, and taking the particle with the minimum fitness as the optimal solution; Taking the minimum positioning error as the search strategy to initialize the particle population, and the expression is: ； where the position of the z-th particle in the d-th dimension is , the random number from 0 to 1 is , the lower limit of the d-th dimension is , the upper limit of the d-th dimension is ; Updating the particle position, and the expression is: ； where the position of the $z$-th particle in the $d$-th dimension at the $(t + 1)$-th iteration is , the optimal solution is , the random integer from 1 to 2 is , the fitness function value of the optimal solution is , the fitness function value of the $z$-th particle is , the current iteration number is $t$, and the maximum iteration number is , the random numbers from 0 to 1 are respectively 、 ; Updating the particle position to obtain the search position, and the expression is: ； where the search position of the $z$-th particle in the $d$-th dimension at the $(t + 1)$-th iteration is , the position of the $z$-th particle in the $d$-th dimension at the $t$-th iteration is , and the random number between 0 and 1 is ; Adopting an attenuation factor to update the search position to obtain the adaptation position, and the expression is: ； where the search position of the $z$-th particle in the $d$-th dimension at the $t$-th iteration is , the random number from 0 to 1 is , the attenuation factor is , and the adaptation position of the $z$-th particle in the $d$-th dimension at the $(t + 1)$-th iteration is ; Adopting a non-linear self-adaptive adjustment of the attenuation factor, and the expression is: ； where the maximum value of the attenuation factor is , the minimum value of the attenuation factor is , and the attenuation factor for the t-th iteration is ; Adopting the updated attenuation factor to update the adaptation position, and iterating continuously until the positioning error reaches the minimum, otherwise updating the particle and readjusting the attenuation factor.

8. A system for associating positioning information based on multi-source data with a train, for performing the method according to any one of claims 1-7, characterized in that, Including: A data acquisition module: used to acquire the train operation data of multiple data sources and preprocess the train operation data; The train operation data includes historical data and real-time data; the historical data includes train route information, train timetables, and operation line data; the real-time data is WTDS data; A clustering and stringing module: used to perform train number clustering on the train operation data to obtain train running data, and perform train number stringing on the train running data to obtain the train running path; A monitoring and correction module: used to monitor the positioning integrity of the train running path to obtain integrity data, and correct the integrity data according to the real-time data to obtain monitored and corrected data; A calculation and construction module: used to calculate the train positioning according to the monitored and corrected data, and construct an association model between the positioning information and the train according to the train positioning; An optimization and output module: used to optimize the association model between the positioning information and the train according to the positioning error, input the data to be associated into the association model between the positioning information and the train, and output the association result; including: Calculate the loss function of the actual train positioning and the predicted train positioning, and take the minimum value of the loss function as the objective function of the positioning information and the train association model; The positioning information and train association model includes a clustering anomaly recognition algorithm, a causal inference method, and a deep belief network algorithm; The clustering anomaly recognition algorithm groups the input multi-source train operation data into different clusters, identifies the data points in the multi-source train operation data that are significantly different from the main cluster, and marks the significantly different data as abnormal data; The causal inference method determines the causes and consequences of anomalies by analyzing the causal relationship between abnormal data and train positioning variables, quantifies the impact of anomalies on the system or dataset based on the causes and consequences, and obtains the impact coefficient; The deep belief network algorithm pre-trains and fine-tunes the deep neural network layer by layer with multi-source train operation data and the impact coefficient, learns complex non-linear relationships, and predicts train positioning information.

Images