A method and system for predicting the running curve of urban rail trains on complex lines
An improved predictive function control algorithm combining wavelet neural networks and multi-mass models has solved the problem of tracking control of urban rail train operation curves under complex lines, achieving high-precision and low-energy-consumption tracking control.
Patent Information
- Application Number
- CN202311280114.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-28
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-09-28
AI Technical Summary
Tracking control of urban rail trains on complex lines is difficult to meet real-world requirements, and traditional algorithms are ineffective in mitigating the impact of track undulations on control performance.
An improved predictive function control algorithm based on wavelet neural network is adopted. By nonlinear adaptive online adjustment of the softening factor and calculation of unit running resistance by multi-mass model, combined with step and wavelet basis function switching, the control parameters are optimized to achieve high-precision and high-comfort tracking control.
It significantly improves the control quality of urban rail trains on complex lines, enhances energy efficiency, punctuality, stopping accuracy and comfort, reduces speed jitter, and improves the accuracy of tracking control.
Smart Images

Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention relates to the field of urban rail train operation tracking and control technology, and more particularly to a method and system for predicting the operation curve of urban rail trains on complex lines. Background Technology
[0002] The difficulty of tracking and controlling urban rail trains on complex tracks stems from the unpredictable undulations of the tracks and the constraints of multiple speed-limited sections. This is an unavoidable and difficult-to-solve practical engineering problem in urban rail transit. Traditional tracking and control algorithms are insufficient to effectively mitigate the impact of complex tracks on the tracking and control performance of urban rail trains. Summary of the Invention
[0003] To address the aforementioned technical problems, this invention provides a method and system for predicting the running curve of urban rail trains on complex lines. Specifically, it proposes an improved predictive function control algorithm for urban rail train running curves based on wavelet neural networks, applicable to complex lines. The proposed algorithm uses step functions and wavelet functions as basis functions, switching them according to the degree of change in the target speed curve. It also presents a nonlinear adaptive online adjustment strategy for the softening factor based on system performance fuzzy satisfaction and optimization factors. The unit running resistance is calculated based on a multi-mass model, and key control parameters are tuned using wavelet neural networks. This enables low-power, high-comfort, high-punctuality, and high-precision stopping tracking control of urban rail trains on complex lines.
[0004] The technical means employed in this invention are as follows: A method for predicting the train operation curve on complex urban rail lines includes the following steps: Step 1: Construct a multi-objective evaluation model for urban rail train tracking and control; Step 2: Based on the model constructed in Step 1, design an improved prediction model for urban rail train operation under complex lines. Specifically, this includes: selecting basis functions; designing reference trajectories based on predicted and process output values; optimizing the reference trajectory and simplifying the objective function through rolling optimization and feedback correction; Step 3: Adjust the softening factor in the improved prediction model online using the following formula:
[0005] In the formula, and These are the maximum and minimum values of the softening factor, respectively. for Additional resistance per unit at location, This is based on the line's flexibility factor increment. and These are the amplification factor for the increment and the optimization factor for the nonlinear increasing trend, respectively, considering the fluctuating conditions of the line. Step 4: Construct a neural network model, collect the operating parameters of urban rail trains on complex lines, and, under the premise of collecting a certain number of control parameter sets and their control effects, train the wavelet neural network. After training, implement offline tuning of control parameters based on the wavelet neural network.
[0006] Furthermore, step 1 specifically includes the following steps: Step 1.1: Based on the force analysis of the train during operation, establish a train dynamics model; Step 1.2: Establish a second-order train model; Step 1.3: Taking into account the preset performance indicators, establish a comprehensive performance evaluation model for the operation of urban rail trains.
[0007] Furthermore, in step 2, the urban rail train tracking control uses a step basis function in the constant speed section and the small gradient coasting section; and a wavelet basis function with tight local support characteristics and flexible layout in the traction section, braking section, and long gradient coasting section; the switching threshold between the step and wavelet basis functions is specified. The calculation expression is shown below.
[0008]
[0009] In the formula, To control the cycle, This is referred to as the controlled interval.
[0010] Furthermore, in step 2, designing the reference trajectory based on the predicted output value and the process output value specifically involves selecting a curve that asymptotically tends towards a future set value based on the predicted output value and the process output value. For an asymptotically stable system, a first-order exponential function is selected as the reference trajectory. The reference trajectory for the prediction function control algorithm is specifically as follows:
[0011] In the formula, The smoothing factor for the reference trajectory. , , The sampling period is The expected rise time constant of the reference trajectory. It is the natural logarithm; , for Time and The tracking settings at any given time; for Reference trajectory value at time, for The process prediction output at time step for The difference in tracking position at any given moment; , The order of the polynomial fitted to the set value. Fitting the polynomial to the set value The coefficient of the order.
[0012] Furthermore, in step 2, after the prediction model is selected, the system output can be directly predicted based on the system input. In discrete state-space form, the prediction model output... The representation of is shown below.
[0013]
[0014] In the formula, , , The coefficient matrix of the state equation of the prediction model. for The state vector of the prediction model at time step [time]. for Input control quantities for the timing system; No. Prediction model output at time step The calculation formula is shown below.
[0015]
[0016] In the formula, free response Forced response ; basis functions Model output under the influence; make , , The objective function can be simplified to: (This can be achieved through rolling optimization and feedback correction)
[0017] because , ,in, , ; make ,but ; In summary, the specific formula for calculating the control quantity in predictive function control can be given, the first... The input control quantity of the timing system is:
[0018] In the formula, , , , .
[0019] Furthermore, in step 3, if the optimization factor Take 0 or Pick Then, the following formula is used to adjust the softening factor online:
[0020] In the formula, and These are the maximum and minimum values of the softening factor, respectively. The softening factor gain coefficient is derived from... and Joint decision, ; The online adjustment function curve of the softening factor is an optimization factor that determines the shape of the online adjustment function curve of the softening factor. It is the natural logarithm.
[0021] Furthermore, in step 3, to more accurately calculate the currently suitable softening factor online, a method for calculating unit additional resistance based on a multi-mass model is used, specifically including the following steps: A multi-mass model is constructed, which treats the train as a chain of multiple mass points, and the force analysis of the train is performed. Based on the train's operating status, the basic resistance and additional resistance are calculated.
[0022] Specifically, the calculation of train resistance includes: Unit basic resistance Calculation:
[0023] In the formula, , , These are parameters related to vehicle type; Calculation of additional resistance of the ramp: Unit slope resistance Numerically approximating the reduced thousandths of the slope where the particle is located. (per mille) (Reconstructed value).
[0024]
[0025] In the formula, These are parameters related to the line; The increased frictional loss caused by a train entering a curved track is called curve resistance, and it is a function of the curve radius. Unit curve resistance. It is inversely proportional to the curve radius.
[0026]
[0027] In the formula, These are parameters related to the line. The radius of the curve; When a train enters a tunnel, the air pressure difference between the front and rear of the train creates resistance that hinders its movement. This resistance is called tunnel additional resistance, and the unit additional resistance is calculated using the following empirical formula:
[0028] In the formula, These are parameters related to the tunnel, primarily concerning the tunnel's cross-sectional area; This refers to the tunnel length; The train will not cross three or more different gradients, curves, or tunnels simultaneously. The specific formulas for calculating the additional resistance from gradients, curves, and tunnels are as follows:
[0029]
[0030]
[0031] In the formula, , , The points are respectively the positions of the mass at the front of the car. Additional resistance from ramps, curves, and tunnels at the location; , , These are the locations of the last different ramps, curves, and tunnels traversed by the front mass of the train; The length of the urban rail vehicle; , , The position of the mass point at the front of the car The unit additional resistance of the slope, curve, and tunnel where the location is situated; , , The position of the mass point at the rear of the vehicle The unit additional resistance of the slope, curve, and tunnel where the location is situated.
[0032] Furthermore, in the wavelet neural network, the activation function is the Morlet wavelet. In wavelet neural networks, the calculation formulas for the hidden layer activation function output and the final output are shown below.
[0033]
[0034] In the formula, and These are the scaling factor and the translation factor, respectively. The expression for calculating the target output function of the selected wavelet neural network is shown below:
[0035] In the formula, This is the expected output; The wavelet neural network is trained using gradient descent. The specific training iterative formula for the weight parameters is shown below: .
[0036] This invention also discloses a prediction system based on the above-mentioned method for predicting the running curve of urban rail trains on complex lines, comprising: The module for constructing a multi-objective evaluation model for urban rail train tracking and control is used to establish a comprehensive performance evaluation model for urban rail train operation based on the acquired dynamic data of the train during operation and preset performance indicators. An improved prediction model construction module is used to perform prediction strategies based on step basis functions and wavelet basis functions for different travel segments of urban rail trains in the constructed multi-objective evaluation model for urban rail train tracking control. The improvement also includes optimizing the reference trajectory involving the softening factor of the predicted output value and the process output value. The online adjustment module for the softening factor is used to adjust the softening factor in the improved prediction model online. The neural network model building module is used to train and construct a neural network based on the obtained operating parameters of urban rail trains on complex lines.
[0037] Compared with the prior art, the present invention has the following advantages: This invention transforms a complex control problem into an approximately equivalent multi-objective optimization problem, which is then solved using an appropriate optimization method. It is applicable to the improved WNN-IPFC algorithm for speed curve prediction function control of urban rail trains based on wavelet neural networks for complex lines, and its specific improvements are described below.
[0038] (1) Based on the analysis of the influence of line conditions on the system response speed in predictive function control, a nonlinear adaptive online adjustment calculation function of the softening factor based on the fuzzy satisfaction of system performance and the optimization factor is further given, which takes into account the unit additional resistance of the line. By adding the optimization factor, the tracking control performance can be further improved.
[0039] (2) In order to calculate the suitable softening factor online more accurately, a method for calculating the unit additional resistance based on a multi-mass model is given, and a control parameter tuning method based on wavelet neural network is adopted.
[0040] To verify the effectiveness of the proposed algorithm, a hardware-in-the-loop experiment was conducted on the tracking control of urban rail trains on complex lines. The results show that, compared with the effectively improved predictive function control algorithm IPFC, the proposed algorithm WNN-IPFC significantly improves the control quality. Attached Figure Description
[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0042] Figure 1 This is a schematic diagram of the force analysis of urban rail trains under the multi-mass model of this invention.
[0043] Figure 2 This is a diagram of the wavelet network structure of the present invention.
[0044] Figure 3 This is a schematic diagram of the target speed, speed limit, and gradient curve in an embodiment of the present invention.
[0045] Figure 4 This is a schematic diagram of the driving trajectory in an embodiment of the present invention.
[0046] Figure 5 This is a curve showing the incremental increase of the softening factor based on the absolute value of the unit additional resistance in this invention.
[0047] Figure 6 These are velocity-distance curves obtained using different algorithms of this invention.
[0048] Figure 7 The graphs show the speed error-time curves obtained by different algorithms of this invention.
[0049] Figure 8 These are time-distance curves obtained by different algorithms of this invention. Detailed Implementation
[0050] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0051] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0052] This invention discloses a method for predicting the running curve of urban rail trains on complex lines, comprising the following steps: Step 1: Construct a multi-objective evaluation model for urban rail train tracking and control; Step 2: Based on the model constructed in Step 1, design an improved prediction model for urban rail train operation under complex lines. Specifically, this includes: selecting basis functions; designing reference trajectories based on predicted and process output values; optimizing the reference trajectory and simplifying the objective function through rolling optimization and feedback correction; Step 3: Adjust the softening factor in the improved prediction model online using the following formula:
[0053] In the formula, and These are the maximum and minimum values of the softening factor, respectively. for Additional resistance per unit at location, This is based on the line's flexibility factor increment. and These are the amplification factor and the nonlinear increasing trend of the increment, respectively, considering the fluctuating conditions of the line. The train tracking control problem on the complex line is transformed into an online adjustment optimization problem of the softening factor with multiple decision variables, which can be effectively solved by intelligent optimization methods. Step 4: Construct a neural network model, collect the operating parameters of urban rail trains on complex lines, and, under the premise of collecting a certain number of control parameter sets and their control effects, train the wavelet neural network. After training, implement offline tuning of control parameters based on the wavelet neural network.
[0054] Step 1 specifically includes the following steps: Step 1.1: Based on the force analysis of the train during operation, establish a train dynamics model; Step 1.2: Establish a second-order train model; Step 1.3: Taking into account the preset performance indicators, establish a comprehensive performance evaluation model for the operation of urban rail trains.
[0055] In this embodiment, the comprehensive performance evaluation model for urban rail train tracking control is specifically designed to make the tracking control performance evaluation indicators as ideal as possible, while also taking into account the optimization of four performance indicators: energy saving, punctuality, stopping accuracy, and comfort.
[0056] In the formula, The integral of time multiplied by the absolute value of the error is a commonly used performance evaluation index for tracking control. , , , These are energy consumption, comfort, absolute value of parking error, and absolute value of punctuality error, respectively. , and For train operation The input control quantity at any given time, the actual operating speed, and the auxiliary system power. and These represent the coefficients for the conversion between electrical energy and mechanical energy during train traction and the conversion between mechanical energy and electrical energy during train braking, respectively. For the train at the Acceleration at each operating point This represents the number of operating points.
[0057] Ideally, urban rail train tracking control aims for the actual speed curve to perfectly match the target speed curve, which means obtaining the integral of time multiplied by the absolute value of the error. But in reality, it's almost impossible. Generally speaking, The smaller the value, the better the tracking control performance, and the lower the energy consumption, comfort, parking accuracy, and on-time error during operation. However, there is also a certain degree of contradiction among them. In practice, under the premise of implementing tracking control smoothly, accurately, and stably, the importance of each indicator should be taken into account in order to ultimately obtain a tracking control result that best meets the control expectations.
[0058] In step 2, the urban rail train tracking control uses a step basis function in the constant speed section and the small gradient coasting section; and a wavelet basis function with tight local support characteristics and flexible layout in the traction section, braking section, and long gradient coasting section. The switching threshold between the step and wavelet basis functions is specified. The calculation expression is shown below.
[0059]
[0060] In the formula, To control the cycle, This is referred to as the controlled interval, with units of velocity, time, and acceleration in km / h, s, and m / s, respectively. 2 The conditions are as a premise.
[0061] In step 2, designing the reference trajectory based on the predicted output value and the process output value specifically involves selecting a curve that asymptotically tends towards the future set value based on the predicted output value and the process output value. For an asymptotically stable system, a first-order exponential function is selected as the reference trajectory. The specific reference trajectory for the prediction function control algorithm is as follows:
[0062] In the formula, The smoothing factor for the reference trajectory. , , The sampling period is The expected rise time constant of the reference trajectory. It is the natural logarithm; , for Time and The tracking settings at any given time; for Reference trajectory value at time, for The process prediction output at time step for The difference in tracking position at any given moment; , The order of the polynomial fitted to the set value. Fitting the polynomial to the set value The coefficient of the order.
[0063] In step 2, after the prediction model is selected, the system output can be directly predicted based on the system input. In discrete state-space form, the prediction model output... The representation of is shown below.
[0064]
[0065] In the formula, , , The coefficient matrix of the state equation of the prediction model. for The state vector of the prediction model at time step [time]. for Input control quantities for the timing system; No. Prediction model output at time step The calculation formula is shown below.
[0066]
[0067] In the formula, free response Forced response ; basis functions Model output under the influence; make , , The objective function can be simplified to: (This can be achieved through rolling optimization and feedback correction)
[0068] because , ,in, , ; make ,but ; In summary, the specific formula for calculating the control quantity in predictive function control can be given, the first... The input control quantity of the timing system is:
[0069] In the formula, , , , .
[0070] In step 3, if the optimization factor Take 0 or Pick Then, the following formula is used to adjust the softening factor online:
[0071] In the formula, and These are the maximum and minimum values of the softening factor, respectively. The softening factor gain coefficient is derived from... and Joint decision, ; The online adjustment function curve of the softening factor is an optimization factor that determines the shape of the online adjustment function curve of the softening factor. It is the natural logarithm.
[0072] In step 3, the multi-mass model describes the force calculation of the train using a mass chain approach, treating the train as a chain of multiple point masses. A schematic diagram of the force analysis of a specific multi-mass model for urban rail trains is shown below. Figure 1 As shown. Figure 1 middle, , , and Represents the first mass of a multi-mass model urban rail train. i The velocity, traction force, support force, gravity, and resistance force of a point mass. Figure 1 As shown, the particle is in a traction state because it is subjected to a traction force in the same direction as its movement; if it is subjected to a braking force in the opposite direction to its movement, it is in a braking state.
[0073] To more accurately calculate the suitable softening factor online, a method for calculating unit additional resistance based on a multi-mass model is used, which includes the following steps: A multi-mass model is constructed, which treats the train as a chain of multiple mass points, and the force analysis of the train is performed. The operating resistance of urban rail trains can be categorized into basic resistance based on its causes. and additional resistance .
[0074] Basic resistance is the resistance that exists under all conditions during train operation. Basic resistance varies with train speed; at low speeds, bearing resistance accounts for a relatively large proportion; as speed increases, the proportions of wheel-rail sliding resistance, impact vibration resistance, and air resistance gradually increase. It is generally measured using unit resistance, which is the resistance experienced per unit weight of the vehicle.
[0075] Specifically, the calculation of train resistance includes: Unit basic resistance Calculation:
[0076] In the formula, , , Additional resistance is a parameter related to the type of train. It is the resistance that a train generates under specific conditions during operation. Additional resistance mainly depends on track conditions. There are various types of additional resistance, such as gradient additional resistance, curve additional resistance, and tunnel additional resistance. These types of additional resistance sometimes exist alone, and sometimes multiple types coexist. When a train is running on a slope, the component of gravity along the track direction is the additional resistance caused by the slope. The calculation of additional resistance caused by the slope is as follows: Unit slope resistance Numerically approximating the reduced thousandths of the slope where the particle is located. (per mille) (Reconstructed value).
[0077]
[0078] In the formula, These are parameters related to the circuit, and are usually close to 1; The increased frictional loss caused by a train entering a curved track is called curve resistance, and it is a function of the curve radius. Unit curve resistance. It is inversely proportional to the curve radius.
[0079]
[0080] In the formula, These are parameters related to the line, typically ranging from 450 to 800; The radius of the curve; When a train enters a tunnel, the air pressure difference between the front and rear of the train creates resistance that hinders its movement. This resistance is called tunnel additional resistance, and the unit additional resistance is calculated using the following empirical formula:
[0081] In the formula, These are parameters related to tunnels, mainly related to the tunnel cross-sectional area, and are generally taken as 0.00013. This refers to the tunnel length; The single-mass train model treats changes in additional resistance abruptly, without a gradual process. This does not reflect the changing pattern of additional resistance as a train gradually moves from one gradient to another. A multi-mass train model, however, can take into account the train's length, thus better handling gradient changes and calculating more accurate additional resistance.
[0082] The train will not cross three or more different gradients, curves, or tunnels simultaneously. The specific formulas for calculating the additional resistance from gradients, curves, and tunnels are as follows:
[0083]
[0084]
[0085] In the formula, , , The points are respectively the positions of the mass at the front of the car. Additional resistance from ramps, curves, and tunnels at the location; , , These are the locations of the last different ramps, curves, and tunnels traversed by the front mass of the train; The length of the urban rail vehicle; , , The position of the mass point at the front of the car The unit additional resistance of the slope, curve, and tunnel where the location is situated; , , The position of the mass point at the rear of the vehicle The unit additional resistance of the slope, curve, and tunnel where the location is situated.
[0086] For practical engineering applications like urban rail train tracking and control, computationally efficient algorithms with low computational cost and high optimization performance are highly favored. Neural networks are lightweight, cost-effective predictive optimization algorithms that require only a small number of control parameter sets and their control effects for initial training—typically 10-20 sets are sufficient. The trained neural network can then be used for offline prediction to quickly obtain control parameter sets with satisfactory control performance. Wavelet neural networks are network models that utilize wavelet bases as the nonlinear activation functions of neurons. Compared to traditional neural networks such as BP and RBF, they possess stronger learning capabilities and more generalized structural capabilities, resulting in superior global optimization quality. The specific wavelet network structure diagram of this invention is shown below. Figure 2 As shown.
[0087] In wavelet neural networks, the activation function chosen is the Morlet wavelet, and its specific expression is shown in the following formula:
[0088] In wavelet neural networks, the calculation formulas for the hidden layer activation function output and the final output are shown below.
[0089]
[0090] In the formula, and These are the scaling factor and the translation factor, respectively. The expression for calculating the target output function of the selected wavelet neural network is shown below:
[0091] In the formula, This is the expected output; The wavelet neural network is trained using gradient descent. The specific training iterative formula for the weight parameters is shown below: .
[0092] By collecting a certain number of control parameter sets and their control effects, a wavelet neural network can be trained. After training, control parameter tuning based on the wavelet neural network can be achieved offline.
[0093] Example 1 The main parameters of the urban rail train operation curve tracking control scenario for complex lines selected in this embodiment are shown in Table 1. The speed limit, target speed, gradient curve, and train trajectory diagram are shown below. Figure 3 , 4 As shown: Table 1. Operation curves of urban rail trains on complex lines
[0094] Based on experience and experimental results, this paper presents the main parameters of the proposed improved predictive function control algorithm. Specifically: the control period is 0.5s, the optimization time domain is 10, and the basis function switching threshold is... The speed is 5.67 km / h, the fitting points are 3, 5, 8, and 9, and the fuzzy width is... and Both were 0.006s. and Both are 0.02 km / h, and the system rise time is the upper and lower limits of the expected design values for the predicted output. and , and The times were 0.008s and 0.02s, respectively. 0.02km / h and 0.02km / h The target velocity is determined by the following parameters. Two different Morlet wavelet basis functions and scaling factors were selected at fitting points 3, 5 and 8, 9, respectively. and The shift factors are 1 and 2 respectively. and The bandwidth parameters are 0 and 5 respectively. and All are 0.5. The optimizer used in this paper incorporates a wavelet neural network algorithm for tuning the control parameters, and the optimization tuning yields 5 parameters, including the maximum value of the softening factor under straight-line conditions. Minimum value Optimization factors , and The values are 0.83, 0.75, 0.92, 0.046, and 0.30, respectively. The specific incremental curves of the softening factor based on the absolute value of unit additional resistance are shown below. Figure 5 As shown.
[0095] To verify the effectiveness of the proposed improved predictive function control algorithm, urban rail train speed curve tracking control was implemented using fuzzy proportional-integral-derivative control (Fuzzy PID), traditional dynamic matrix control (TDMC), improved predictive function control (IPFC), and the wavelet neural network-based improved predictive function control algorithm WNN-IPFC. Fuzzy PID and TDMC are two widely used traditional improved algorithms, while IPFC is an effective improved predictive function control algorithm. Specific simulation results for urban rail train speed curve tracking control are shown in Table 2 and... Figure 6-8 As shown.
[0096] Table 2 Tracking control results obtained by different algorithms Table 2 Tracking control results of different algorithms
[0097] As shown in Table 2, based on IPFC, WNN-IPFC can further and effectively improve control quality, including four performance indicators: energy saving, punctuality, precise stopping, and comfort, as well as the time multiplied by the absolute value of the error integral. All have shown considerable improvement; compared to IPFC, the four indicators of energy saving, punctuality, precise parking, and comfort have improved by 2.1%, 25.4%, 43.4%, and 18.0%, respectively. It increased by 23.7%.
[0098] Depend on Figure 6 and Figure 8 It can be seen that, based on IPFC, WNN-IPFC can obtain a more ideal actual tracking control curve; from Figure 7 It can be seen that, based on IPFC, WNN-IPFC further significantly suppresses speed chattering during the tracking control process; Figure 8 It can be seen that, based on IPFC, WNN-IPFC obtained a more ideal actual parking point, which is closer to the target parking point (198s, 2670m).
[0099] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0100] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0101] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.
[0102] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0103] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0104] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting the running curve of urban rail trains on complex lines, characterized in that, Includes the following steps: Step 1: Construct a multi-objective evaluation model for urban rail train tracking and control; Step 2: Based on the model constructed in Step 1, design an improved prediction model for urban rail train operation under complex lines. Specifically, this includes: selecting basis functions; designing reference trajectories based on predicted and process output values; optimizing the reference trajectory and simplifying the objective function through rolling optimization and feedback correction; Step 3: Adjust the softening factor in the improved prediction model online using the following formula: In the formula, and These are the maximum and minimum values of the softening factor, respectively. for Additional resistance per unit at location, This is based on the line's flexibility factor increment. and These are the amplification factor for the increment and the optimization factor for the nonlinear increasing trend, respectively, considering the fluctuating conditions of the line. Step 4: Construct a neural network model, collect the operating parameters of urban rail trains on complex lines, and, under the premise of collecting a certain number of control parameter sets and their control effects, train the wavelet neural network. After training, implement offline tuning of control parameters based on the wavelet neural network. In step 2, the urban rail train tracking control uses a step basis function in the constant speed section and the small gradient coasting section; and a wavelet basis function with tight local support characteristics and flexible layout in the traction section, braking section, and long gradient coasting section. The switching threshold between the step and wavelet basis functions is specified. The calculation expression is as follows: In the formula, To control the cycle, This is called the controlled interval; In step 3, to more accurately calculate the suitable softening factor online, a method for calculating unit additional resistance based on a multi-mass model is used, which specifically includes the following steps: A multi-mass model is constructed, which treats the train as a chain of multiple mass points, and the force analysis of the train is performed. Based on the train's operating status, the basic resistance and additional resistance are calculated; Specifically, the calculation of train resistance includes: Unit basic resistance Calculation: In the formula, , , These are parameters related to vehicle type; Calculation of additional resistance of the ramp: Unit slope resistance Numerically approximating the reduced thousandths of the slope where the particle is located. ; In the formula, These are parameters related to the line; The increased frictional loss caused by a train entering a curved track results in resistance called curve-additional resistance, which is a function of the curve radius: unit curve resistance. Inversely proportional to the curve radius: In the formula, These are parameters related to the line. The radius of the curve; When a train enters a tunnel, the air pressure difference between the front and rear of the train creates resistance that hinders its movement. This resistance is called tunnel additional resistance, and the unit additional resistance is calculated using the following empirical formula: In the formula, These are parameters related to the tunnel, primarily concerning the tunnel's cross-sectional area; This refers to the tunnel length; The train will not cross three or more different gradients, curves, or tunnels simultaneously. The specific formulas for calculating the additional resistance from gradients, curves, and tunnels are as follows: In the formula, , , The points are respectively the positions of the mass at the front of the car. Additional resistance from ramps, curves, and tunnels at the location; , , These are the locations of the last different ramps, curves, and tunnels traversed by the front mass of the train; The length of the urban rail vehicle; , , The position of the mass point at the front of the car The unit additional resistance of the slope, curve, and tunnel where the location is situated; , , The position of the mass point at the rear of the vehicle The unit additional resistance of the slope, curve, and tunnel where the location is situated; In wavelet neural networks, the Morlet wavelet is chosen as the activation function. In wavelet neural networks, the formulas for calculating the hidden layer activation function output and the final output are as follows: In the formula, and These are the scaling factor and the translation factor, respectively. The expression for calculating the target output function of the selected wavelet neural network is shown below: In the formula, This is the expected output; The wavelet neural network is trained using gradient descent. The specific training iterative formula for the weight parameters is shown below: 。 2. The method for predicting the running curve of urban rail trains on complex lines according to claim 1, characterized in that, Step 1 specifically includes the following steps: Step 1.1: Based on the force analysis of the train during operation, establish a train dynamics model; Step 1.2: Establish a second-order train model; Step 1.3: Taking into account the preset performance indicators, establish a comprehensive performance evaluation model for the operation of urban rail trains.
3. The method for predicting the running curve of urban rail trains on complex lines according to claim 1, characterized in that, In step 2, designing the reference trajectory based on the predicted output value and the process output value specifically involves selecting a curve that asymptotically tends towards the future set value based on the predicted output value and the process output value. For an asymptotically stable system, a first-order exponential function is selected as the reference trajectory. The specific reference trajectory for the prediction function control algorithm is as follows: In the formula, The smoothing factor for the reference trajectory. , , The sampling period is The expected rise time constant of the reference trajectory. It is the natural logarithm; , for Time and The tracking settings at any given time; for Reference trajectory value at time, for The process prediction output at time step for The difference in tracking position at any given moment; , The order of the polynomial fitted to the set value. Fitting the polynomial to the set value The coefficient of the order.
4. The method for predicting the running curve of urban rail trains on complex lines according to claim 3, characterized in that, In step 2, after the prediction model is selected, the system output can be directly predicted based on the system input. In discrete state-space form, the prediction model output... The representation of is as follows: In the formula, , , The coefficient matrix of the state equation of the prediction model. for The state vector of the prediction model at time step [time]. for Input control quantities for the timing system; No. Prediction model output at time step The calculation formula is as follows: In the formula, free response Forced response ; basis functions Model output under the influence; make , , The objective function can be simplified to: (This can be achieved through rolling optimization and feedback correction) because , ,in, , ; make ,but ; In summary, the specific formula for calculating the control quantity in predictive function control can be given, the first... The input control quantity of the timing system is: In the formula, , , , 。 5. The method for predicting the running curve of urban rail trains on complex lines according to claim 1, characterized in that, In step 3, if the optimization factor Take 0 or Pick Then, the following formula is used to adjust the softening factor online: In the formula, and These are the maximum and minimum values of the softening factor, respectively. The softening factor gain coefficient is derived from... and Joint decision, ; The online adjustment function curve of the softening factor is an optimization factor that determines the shape of the online adjustment function curve of the softening factor. It is the natural logarithm.
6. A prediction system based on the method for predicting the running curve of urban rail trains under complex lines as described in any one of claims 1 to 5, characterized in that, include: The module for constructing a multi-objective evaluation model for urban rail train tracking and control is used to establish a comprehensive performance evaluation model for urban rail train operation based on the acquired dynamic data of the train during operation and preset performance indicators. An improved prediction model construction module is used to perform prediction strategies based on step basis functions and wavelet basis functions for different travel segments of urban rail trains in the constructed multi-objective evaluation model for urban rail train tracking control. The improvement also includes optimizing the reference trajectory involving the softening factor of the predicted output value and the process output value. The online adjustment module for the softening factor is used to adjust the softening factor in the improved prediction model online. The neural network model building module is used to train and construct a neural network based on the obtained operating parameters of urban rail trains on complex lines.
Citation Information
Patent Citations
Subway traffic flow optimization control method
CN105083335A
Method and system for controlling heavy-haul train based on reinforcement learning
US11205124B1
