A train operation control optimization method based on time-varying parameter identification

Abstract

The present invention discloses a train operation control optimization method based on time-varying parameter identification, comprising the following steps: step S1, determining input and output data through the records of on-board equipment during train operation, analyzing the time-varying parameters that the train needs to identify, and establishing a basic train model; step S2, obtaining a train operation model based on the auxiliary model through auxiliary model processing for actual collected data; step S3, using the train operation model obtained in step S2, applying it to the train control system, and optimizing the train operation. The present invention is based on data-driven parameter identification, starting from actual input and output data, taking into account the variability of the actual train operation environment, by continuously iteratively updating information, and accelerating the decay of historical data with a forgetting factor, it can track environmental changes well, realize the identification of time-varying parameters, have lower prediction errors, and can achieve optimization of train control.

Description

Technical Field

[0001] The invention relates to the technical field of rail transportation, and in particular to a train operation control optimization method based on time-varying parameter identification. Background Art

[0002] Due to its outstanding performance in energy saving and relieving traffic congestion, urban rail transit has developed rapidly in major cities. By the end of 2020, a total of 44 cities in mainland China had opened urban rail transit lines, with a total of 233 lines and an operating mileage of 7,545.5 kilometers. In 2020, 39 new rail transit lines were added throughout the year, with an additional operating mileage of more than 1,240 kilometers, an increase of 20.1% compared with 2019. The development of urban rail transit has led to the rapid development of the railway construction industry, urban rail train design and manufacturing industry, and signal equipment suppliers. The research on scientific and technological means in the rail transit industry has been ongoing to achieve better rail transit goals.

[0003] Train operation can be regarded as a constantly changing process. The operation data of the train also increases over time, and there will be randomly changing noise interference in the process, so the train operation system can be regarded as a time-varying system. For the traditional train operation model, every time new train data is generated, all the data needs to be recalculated, which takes a lot of time and is not very practical. The existing technology relies more on some ideal assumptions and empirical formula parameters, and the model of a simple single particle is still established. The actual parameters of the train are indeed difficult to obtain, resulting in the simplification of the actual train parameters, line parameters, etc. in the actual processing. Therefore, it is necessary to adopt a method that can identify time-varying parameters, construct a train operation model, and optimize the train operation control. Summary of the invention

[0004] In response to the above-mentioned shortcomings in the prior art, the applicant provides a reasonable and effective train operation control optimization method based on time-varying parameter identification, which can realize online identification of train operation parameters. By continuously iteratively updating information, it can track environmental changes well, has lower prediction error, and can achieve optimization of train control.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A train operation control optimization method based on time-varying parameter identification comprises the following steps:

[0007] Step S1, through the records of onboard equipment during train operation, determine the input and output data, analyze the time-varying parameters that need to be identified in the train, and establish a basic train model, specifically including:

[0008] Formula (1) for establishing the train longitudinal dynamic equation is:

[0009]

[0010] Among them, ε is the acceleration coefficient, and α0, α1, and α2 are all time-varying parameters;

[0011] By performing differential transformation on formula (1) and introducing random interference noise δ(k) in the actual environment, the differential equation formula (2) is obtained:

[0012] y(k)=a1(k)·x(k-1) 2 +a2(k)·x(k-1)+b1(k)·u(k-1)+b2(k)+δ(k) (2);

[0013] Among them, x(k) is the real speed of the train, u(k) is the traction / braking force on the train, a1, a2, b1, b2 are all time-varying parameters that need to be identified, and k is time; y(k) is the actual measured train speed in the presence of noise, which is the sum of the real speed x(k) output by the train system and the random interference noise δ(k), that is:

[0014] y(k)=x(k)+δ(k) (3);

[0015] make

[0016] θ(k)=[a1(k), a2(k), b1(k), b2(k)] T (4);

[0017]

[0018] get:

[0019]

[0020]

[0021] Among them, θ(k) is the parameter vector, is the information vector. Definition is the parameter estimate calculated based on the first K groups of data, then the prediction error ε(k) at the kth measurement is:

[0022]

[0023] Therefore, the sum of the squares of the difference between the actual output of the system and the estimated output of the model, that is, the criterion function is:

[0024]

[0025] Formula (9) calculates input and output based on the new data continuously collected during the train operation, thereby updating the estimated values ​​of the train parameters. The specific calculation formula is as follows:

[0026]

[0027] Where K(k) is the Kalman gain, P(k) is the covariance matrix of the error, and λ is the forgetting factor;

[0028] Step S2, obtaining a train operation model based on the auxiliary model by processing the actual collected data, specifically includes:

[0029] Build an output error model:

[0030]

[0031] The auxiliary model is established as:

[0032]

[0033] Among them, x a (k) is the output of the auxiliary model, is the estimated value of the information vector, is the estimated value of the parameter vector; so there is the following relationship:

[0034]

[0035]

[0036] Through the above auxiliary model, the estimated value of the information vector is As the information vector of the auxiliary system, the estimated value of the parameter vector is As the parameter vector of the auxiliary system, the train operation model based on the auxiliary model is obtained as follows:

[0037]

[0038] Step S3, adopting the train operation model obtained in step S2 and applying it to the train control system to optimize the train operation.

[0039] The beneficial effects of the present invention are as follows:

[0040] The present invention determines the input and output data through the records of the on-board equipment during the operation of the train, thereby establishing a model structure. Then, the initial value is specified to construct the information vector, and the actual output data is used as the target. The residual is calculated by subtracting the value calculated by the model from the actual value. During the operation process, new data is continuously added, and the residual is brought into the formula of parameter recursion, and the parameters of the previous round are updated. In this way, continuous iteration begins to achieve parameter update. The implementation of recursion is based on the continuous increase of data. Only one parameter estimation is required each time, and then continuous iterative updates are performed based on the new data. This can well realize the online identification of train operation parameters.

[0041] The present invention is based on data-driven parameter identification. Since it ignores the internal structure of the system and starts from the actual input and output data, it takes into account the variability of the actual train operation environment. It updates information through continuous iteration and accelerates the decay of historical data with a forgetting factor. It can track environmental changes well and realize the identification of time-varying parameters. It has lower prediction errors and can achieve the optimization of train control. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0043] The specific implementation of the present invention will be described below in conjunction with the accompanying drawings.

[0044] The train operation control optimization method based on time-varying parameter identification described in the present invention constructs a train operation model from the perspective of parameter identification, and takes into account the dynamic process of train operation, and adopts a time-varying parameter identification method to complete the identification of the train's time-varying parameters. The present invention performs parameter identification on the operation of urban rail trains, mainly based on a large amount of real-time operation data collected during the train's travel and the knowledge of the train's dynamic equations to determine a set of parameter values. The input here is mainly the traction / braking force of the train, and the output is the running speed of the train, so that the unknown process can be predicted through the determined parameters.

[0045] Refer to Figure 1 As shown, the train operation control optimization method based on time-varying parameter identification of the present invention comprises the following steps:

[0046] Step S1, through the records of onboard equipment during train operation, determine the input and output data, analyze the time-varying parameters that need to be identified, and establish a basic train model. Specifically including:

[0047] Formula (1) for establishing the train longitudinal dynamic equation is:

[0048]

[0049] Among them, ε is the acceleration coefficient, and α0, α1, and α2 are all time-varying parameters. u(t) is the traction / braking force on the train at time t, and v(t) is the speed of the train at time t.

[0050] For train parameter identification, it is generally necessary to transform the model into a differential form, perform differential transformation on formula (1) based on the train dynamics equation, and introduce random interference noise δ(k) in the actual environment to obtain the differential equation formula (2):

[0051] y(k)=a1(k)·x(k-1) 2 +a2(k)·x(k-1)+b1(k)·u(k-1)+b2(k)+δ(k) (2);

[0052] Among them, x(k) is the real speed of the train, u(k) is the traction / braking force on the train, a1, a2, b1, b2 are all time-varying parameters that need to be identified, and k is time. Here y(k) is the actual measured train speed in the presence of noise, which is the sum of the real speed x(k) output by the train system and the random interference noise δ(k), that is:

[0053] y(k)=x(k)+δ(k) (3);

[0054] make

[0055] θ(k)=[a1(k), a2(k), b1(k), b2(k)] T (4);

[0056]

[0057] get:

[0058]

[0059]

[0060] Among them, θ(k) is the parameter vector, is the information vector. Definition is the parameter estimate calculated based on the first K groups of data, then the prediction error ε(k) at the kth measurement is:

[0061]

[0062] Therefore, the sum of the squares of the difference between the actual output of the system and the estimated output of the model, that is, the criterion function J(θ), is:

[0063]

[0064] Formula (9) calculates input and output based on the new data continuously collected during the train operation, thereby updating the estimated values ​​of the train parameters. The specific calculation formula is as follows:

[0065]

[0066] Among them, K(k) is the Kalman gain, P(k) is the covariance matrix of the error, and λ is the forgetting factor.

[0067] Step S2, by processing the auxiliary model for the actual collected data, obtain a train operation model based on the auxiliary model. Specifically including:

[0068] Build an output error model:

[0069]

[0070] The auxiliary model is established as:

[0071]

[0072] Among them, x a (k) is the output of the auxiliary model, and the estimated value of the information vector is The estimated value of the parameter vector is So there is the following relationship:

[0073]

[0074]

[0075] Through the above auxiliary model, the estimated value of the information vector is As the information vector of the auxiliary system, the estimated value of the parameter vector As the parameter vector of the auxiliary system, the train operation model based on the auxiliary model is obtained as follows:

[0076]

[0077] Step S3, adopting the train operation model obtained in step S2 and applying it to the train control system to optimize the train operation.

[0078] The present invention determines the input and output data through the records of the onboard equipment during the train operation, thereby establishing the model structure. Then, the initial value is specified to construct the information vector, and the actual output data is taken as the target. The residual is calculated by subtracting the value calculated by the model from the actual value. During the operation, new data is continuously added, and the residual is brought into the parameter recursive formula to update the parameters of the previous round, thus starting continuous iteration to achieve parameter update.

[0079] The implementation of recursion is based on the ever-increasing data. The parameter estimate of the system at the current moment comes from the parameter estimate at the previous moment plus a certain correction term: That is to say, only one parameter estimation is required each time, and then continuous iterative updates are performed based on new data, which can well realize the online identification of train operation parameters.

[0080] The present invention is based on data-driven parameter identification. Since it ignores the internal structure of the system and starts from the actual input and output data, taking into account the variability of the actual train operation environment, it updates information through continuous iteration and accelerates the decay of historical data with a forgetting factor. It can track environmental changes well, realize the identification of time-varying parameters, have lower prediction errors, and achieve better control effects.

[0081] The above description is an explanation of the present invention, not a limitation of the present invention. The present invention may be modified in any form without violating the spirit of the present invention.

Claims

1. A train operation control optimization method based on time-varying parameter identification, comprising the following steps: Step S1, through the records of onboard equipment during train operation, determine the input and output data, analyze the time-varying parameters that need to be identified in the train, and establish a basic train model, specifically including: Formula (1) for establishing the train longitudinal dynamic equation is: Among them, ε is the acceleration coefficient, and α0, α1, and α2 are all time-varying parameters; By performing differential transformation on formula (1) and introducing random interference noise δ(k) in the actual environment, the differential equation formula (2) is obtained: y(k)=a1(k)·x(k-1) 2 +a2(k)·x(k-1)+b1(k)·u(k-1)+b2(k)+δ(k) (2); Among them, x(k) is the real speed of the train, u(k) is the traction / braking force on the train, a1, a2, b1, b2 are all time-varying parameters that need to be identified, and k is time; y(k) is the actual measured train speed in the presence of noise, which is the sum of the real speed x(k) output by the train system and the random interference noise δ(k), that is: y(k)=x(k)+δ(k) (3); make θ(k)=[a1(k)、a2(k)、b1(k)、b2(k)] T (4); get: Among them, θ(k) is the parameter vector, is the information vector, and we define is the parameter estimate calculated based on the first K groups of data, then the prediction error ε(k) at the kth measurement is: Therefore, the sum of the squares of the difference between the actual output of the system and the estimated output of the model, that is, the criterion function is: Formula (9) calculates input and output based on the new data continuously collected during the train operation, thereby updating the estimated values ​​of the train parameters. The specific calculation formula is as follows: Where K(k) is the Kalman gain, P(k) is the covariance matrix of the error, and λ is the forgetting factor; Step S2, obtaining a train operation model based on the auxiliary model by processing the actual collected data, specifically includes: Build an output error model: The auxiliary model is established as: Among them, x a (k) is the output of the auxiliary model, is the estimated value of the information vector, is the estimated value of the parameter vector; so there is the following relationship: Through the above auxiliary model, the estimated value of the information vector is As the information vector of the auxiliary system, the estimated value of the parameter vector is As the parameter vector of the auxiliary system, the train operation model based on the auxiliary model is obtained as follows: Step S3, adopting the train operation model obtained in step S2 and applying it to the train control system to optimize the train operation.

Images