CN-122009286-A - Position determining method and device based on digital twin magnetic suspension train braking process

CN122009286ACN 122009286 ACN122009286 ACN 122009286ACN-122009286-A

Abstract

The invention discloses a position determining method and device based on a digital twin magnetic suspension train braking process, relates to the technical field of rail transit digital twin, and mainly aims to solve the problem that the parking position cannot be ensured to be accurate based on manual observation and operation in the parking process. The method mainly comprises the steps of constructing a linear and nonlinear combined train position prediction model, predicting the output of an unknown nonlinear dynamic system based on a trained online deep learning prediction model, responding to a train stop preparation instruction, acquiring real-time operation data of a target train, inputting the real-time operation data into the linear model, calculating a linear prediction value, predicting the real-time operation data based on the trained online deep learning prediction model to obtain a nonlinear prediction value, adding the linear prediction value and the nonlinear prediction value to obtain a position prediction value of the target train, and outputting the position prediction value to an interactive terminal. The method is mainly used for determining the train position during braking.

Inventors

LIU LING
SONG BAODONG
LIU YUHENG
WANG YAN
DENG LAN
SONG SHUQI
JIA YAO
WANG JIA
GE LUMING
LIU JIANG

Assignees

北京全路通信信号研究设计院集团有限公司
东北大学

Dates

Publication Date: 20260512
Application Date: 20251226

Claims (10)

1.A method for determining a position of a digital twinning-based magnetic levitation train braking process, comprising: Constructing a linear and nonlinear combined train position forecasting model, wherein the train position forecasting model comprises an offline identification linear model and an unknown nonlinear dynamic system, and the output of the unknown nonlinear dynamic system is obtained by prediction based on a trained online deep learning forecasting model; responding to a train stopping preparation instruction, and acquiring real-time operation data of a target train; inputting the real-time operation data into the linear model, calculating a linear prediction value, and carrying out prediction processing on the real-time operation data based on the trained online deep learning prediction model to obtain a nonlinear prediction value; And adding the linear forecast value and the nonlinear forecast value to obtain a position forecast value of the target train, and outputting the position forecast value to the interactive terminal to assist in realizing stopping to the target parking spot.
2. The method of claim 1, wherein said constructing a linear and nonlinear combined train position forecast model comprises: The method comprises the steps of constructing a train position control closed-loop system equation, decomposing a train position dynamic system into a linear part and a nonlinear part based on the train position control closed-loop system equation, wherein the nonlinear part comprises a plurality of nonlinear fluctuation items used for representing the nonlinear fluctuation caused by uncertain factors in the train control system, identifying parameter vectors of the linear part through a least square algorithm according to historical train operation data to obtain a linear model, constructing an unknown nonlinear dynamic system according to identification errors of the parameter vectors and the unknown nonlinear items, constructing a deep learning model used for predicting the unknown nonlinear dynamic system, and training the deep learning model based on historical train position time sequence data to obtain a trained online deep learning forecast model; And constructing a train position forecasting model based on the linear model, the unknown nonlinear dynamic system and the trained online deep learning forecasting model.
3. The method of claim 2, wherein the train position control closed loop system equation construction process comprises: Constructing a position dynamic model comprising a linear model and an unknown nonlinear term, wherein the linear model comprises a control level and a driving license; Constructing a control law comprising a dynamic error proportion term and a feedforward compensation term, wherein the dynamic error proportion term is dynamically adjusted based on deviation of a real-time position and a target position of a train, and the feedforward compensation term is calculated based on a plurality of train operation state variables, wherein the train operation state variables comprise at least one of a level, a driving license, a strictest target point position, a strictest target point speed, a train speed, a reference speed value, a temporary speed limit, a recommended speed, an emergency braking speed, a highest speed limit, an alarm speed and a gradient; And (3) the output of the control law is acted on the position dynamic model to obtain a train position control closed-loop system equation.
4. The method of claim 2, wherein the identifying the parameter vector of the linear portion by a least squares algorithm based on the historical train operation data to obtain the linear model comprises: constructing an input matrix and an output matrix of train operation control based on historical operation data at continuous moments, wherein the input matrix comprises train positions, target parking points, control levels and train operation state variables at different historical moments; Constructing a parameter identification equation taking the parameter vector of the linear part as an identification object according to the input matrix and the output matrix; And solving the parameter identification equation based on a least square algorithm to obtain an estimated value of the parameter vector, and updating the parameter vector of the linear part by using the estimated value to obtain a linear model.
5. The method of claim 2, wherein the historical train position timing data comprises historical train operation data and historical position measurements for successive moments; the construction and training process of the online deep learning forecast model comprises the following steps: Constructing a deep learning model comprising a multi-layer recurrent neural network, wherein the multi-layer recurrent neural network adopts a framework comprising an input gate, a forgetting gate, an output gate and a state memory unit and is used for extracting nonlinear characteristics of train position time sequence data; calculating the historical linear output of a linear model corresponding to the train historical operation data at each moment, and the difference value between the historical position measured value and the historical linear output, and constructing a training sample set of an initial online deep learning forecasting model by taking the difference value as a sample label of an unknown nonlinear dynamic system; And training the deep learning model by using the training sample set to learn the change rule of the unknown nonlinear dynamic system, so as to obtain a trained online deep learning forecasting model.
6. The method according to any one of claims 1-5, wherein the method is implemented based on an end-to-end cloud architecture, and the acquiring real-time operation data of the target train includes acquiring the real-time operation data based on an end-to-end sensor deployed on the target train and uploading the real-time operation data to a cloud-side real-time operation database; before inputting the real-time operational data into the linear model, the method further comprises: Constructing a linear model on the cloud side and performing online training of an online deep learning forecasting model; The side loads the linear model and the trained online deep learning forecasting model from the cloud side so as to calculate a position forecasting value according to real-time operation data on the side through the linear model and the trained online deep learning forecasting model; The online deep learning forecasting model is on the cloud side and performs online training based on a network parameter self-correction mechanism and a network structure self-correction mechanism.
7. The method according to claim 1, wherein the method further comprises: On the cloud side, respectively carrying out network parameter self-correction and network structure self-correction on the online deep learning forecast model based on the train operation data updated in real time to obtain a network structure self-correction model and a network parameter self-correction model; When the first error of the network parameter self-correction model is monitored to be larger than a first preset error threshold value and the second error of the network structure self-correction model is monitored to be smaller than the first preset error threshold value, correcting the network structure of the network parameter self-correction model based on the network structure of the network structure self-correction model so as to continue on-line training based on the network parameter self-correction model after network structure updating; And under the condition that the third error of the online deep learning forecasting model on the monitored side is larger than a second preset error threshold and the first error of the network parameter self-correction model is smaller than the second preset error threshold, obtaining an updated online deep learning forecasting model based on the weight and the bias of the network parameter self-correction model and the weight and the bias of the online deep learning forecasting model on the offset correction side.
8. A position determining device based on digital twinning for a magnetic levitation train braking process, comprising: the system comprises a construction module, a prediction module and a prediction module, wherein the construction module is used for constructing a linear and nonlinear combined train position prediction model, the train position prediction model comprises an offline identification linear model and an unknown nonlinear dynamic system, and the output of the unknown nonlinear dynamic system is predicted based on a trained online deep learning prediction model; The acquisition module is used for responding to the train stopping preparation instruction and acquiring real-time operation data of the target train; The calculation module is used for inputting the real-time operation data into the linear model to calculate a linear prediction value, and carrying out prediction processing on the real-time operation data based on the trained online deep learning prediction model to obtain a nonlinear prediction value; and the output module is used for adding the linear forecast value and the nonlinear forecast value to obtain a position forecast value of the target train and outputting the position forecast value to the interactive terminal so as to assist in realizing stopping to the target stopping point.
9. A storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the method for determining the position of a digital twinned based maglev train braking process according to any one of claims 1-7.
10. The terminal is characterized by comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; The memory is configured to store at least one executable instruction that causes the processor to perform operations corresponding to the method for determining a position of a digital twinning-based maglev train braking process according to any one of claims 1-7.

Description

Position determining method and device based on digital twin magnetic suspension train braking process Technical Field The invention relates to the technical field of digital twin of rail transit, in particular to a method and a device for determining the position of a magnetic suspension train in a braking process based on digital twin. Background When the magnetic suspension train runs at high speed, the vehicle-mounted speed measuring system (such as a speed measuring device based on an eddy current sensor) can provide relatively accurate speed feedback. However, this manner of speed measurement, which relies on the perception of eddy current sensors, presents a fundamental challenge when the speed of the train is significantly reduced as it enters the end of the braking. For the eddy current sensor, when the speed of the train is reduced after the train enters a braking stage, the eddy current generation amount is reduced, the signal is weak, the signal to noise ratio is low, and the eddy current sensor has larger error in a low-speed state. At this time, the automatic driving system (ATO) estimates the train position based on the misaligned speed signal and controls braking, and a deviation between the train position and the actual position is liable to occur. In view of the fact that the error of an automatic driving system is large in a parking stage, the existing parking process mainly depends on manual work, a driver observes the distance of a reference object of a platform, and a gear handle is manually operated to brake a train to control the speed of the train, so that the accuracy of a parking position is ensured. However, the manual reference object observation method has high uncertainty and unreliability, and the accuracy of the parking position cannot be ensured. Disclosure of Invention In view of the above, the invention provides a method and a device for determining the position of a magnetic suspension train in a braking process based on digital twinning, which mainly aim to solve the problem that the accurate parking position cannot be ensured based on manual observation and operation in the parking process. According to one aspect of the present invention, there is provided a method for determining the position of a digital twin-based magnetic levitation train braking process, comprising: Constructing a linear and nonlinear combined train position forecasting model, wherein the train position forecasting model comprises an offline identification linear model and an unknown nonlinear dynamic system, and the output of the unknown nonlinear dynamic system is obtained by prediction based on a trained online deep learning forecasting model; responding to a train stopping preparation instruction, and acquiring real-time operation data of a target train; inputting the real-time operation data into the linear model, calculating a linear prediction value, and carrying out prediction processing on the real-time operation data based on the trained online deep learning prediction model to obtain a nonlinear prediction value; And adding the linear forecast value and the nonlinear forecast value to obtain a position forecast value of the target train, and outputting the position forecast value to the interactive terminal to assist in realizing stopping to the target parking spot. Further, the constructing a linear and nonlinear combined train position forecasting model includes: The method comprises the steps of constructing a train position control closed-loop system equation, decomposing a train position dynamic system into a linear part and a nonlinear part based on the train position control closed-loop system equation, wherein the nonlinear part comprises a plurality of nonlinear fluctuation items used for representing the nonlinear fluctuation caused by uncertain factors in the train control system, identifying parameter vectors of the linear part through a least square algorithm according to historical train operation data to obtain a linear model, constructing an unknown nonlinear dynamic system according to identification errors of the parameter vectors and the unknown nonlinear items, constructing a deep learning model used for predicting the unknown nonlinear dynamic system, and training the deep learning model based on historical train position time sequence data to obtain a trained online deep learning forecast model; And constructing a train position forecasting model based on the linear model, the unknown nonlinear dynamic system and the trained online deep learning forecasting model. Further, the construction process of the train position control closed loop system equation comprises the following steps: Constructing a position dynamic model comprising a linear model and an unknown nonlinear term, wherein the linear model comprises a control level and a driving license; Constructing a control law comprising a dynamic error proportion term and a feedforward compensation term, wherein