A pantograph active control method considering controlling different components
By designing a controller based on PID, RBF neural network and constrained particle swarm optimization, the problem of contact force fluctuation in the existing technology is solved, and more efficient pantograph active control is achieved, which improves the operation safety and flow of the train. Quality reduces operating costs.
Patent Information
- Application Number
- CN202411073951.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-07
AI Technical Summary
The existing methods to improve the dynamic contact relationship of the bow net are limited and costly, and it is difficult to effectively reduce the fluctuations in the contact force between the pantograph and the contact net, affecting the operation safety and flow quality of the train.
A pantograph active control method considering controlling different components was designed. By establishing a pantograph-contact network system simulation model, a controller based on PID, RBF neural network and constrained particle swarm optimization was designed to realize contact force of the bow network. fine adjustment and control to reduce contact force fluctuations.
It effectively reduces the fluctuation of the contact force of the bow net, improves the flow quality, reduces the operation and maintenance costs, and can still ensure the stability and trajectory tracking performance of the pantograph under the conditions of actuator constraints and time delay.
Smart Images

Figure CN118584821B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of pantograph active control, and in particular relates to a design of a pantograph active control method taking into account the control of different components. Background Art
[0002] Taking rail transit has gradually become a part of the daily travel life of Chinese people. At present, my country's railway trains usually use electric traction to obtain electricity through the pantograph-contact network system to meet the energy supply required for train operation. However, with the gradual increase in train speed and the more complex train operation environment, the coupled contact relationship between the pantograph and the contact network has deteriorated. The gradual increase in train speed and the complex external environment will lead to intensified contact force fluctuations. If the contact force is too large, it will further cause accelerated wear of the pantograph bow head slide, increasing the cost of operation and maintenance. If the contact force is too small, it will increase the probability of the pantograph bow head being offline, further affecting the current collection of the train during operation, greatly endangering the operation safety of the train.
[0003] In recent years, in order to cope with the intensified fluctuations in the pantograph-catenary contact force and the deterioration of the train current-collecting quality, domestic and foreign scholars have conducted a lot of research and analysis on improving the dynamic contact relationship between the pantograph and the catenary, and proposed a series of corresponding solutions, such as optimizing the structural parameters such as the mass, stiffness and damping of the pantograph head and the frame part, improving the contact network structural parameters, and optimizing the distance between the two pantographs of the train. Although the above optimization measures can reduce the fluctuations between the pantograph and the catenary to a certain extent, their effects are still limited. To achieve more significant improvements, it is usually necessary to build new lines or transform existing lines. However, both methods require considerable costs. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that the existing methods for improving the dynamic contact relationship between the pantograph and the catenary have limited effects and high costs, and propose an active pantograph control method that takes into account the control of different components.
[0005] The technical solution of the present invention is: a pantograph active control method considering controlling different components, comprising the following steps:
[0006] S1. Establish a simulation model of the pantograph-catenary system;
[0007] S2. According to the pantograph-contact network system simulation model, the pantograph head and frame are taken as the control objects, and the design is based on PID Controller;
[0008] S3. According to the pantograph-contact network system simulation model, the pantograph head and the upper frame are simultaneously taken as control objects, and the design is based on RBF Adaptive controllers based on neural networks;
[0009] S4. According to the pantograph-catenary system simulation model, the pantograph chassis is taken as the control object, and a predictive controller based on constrained particle swarm optimization is designed;
[0010] S5. Based on PID Controller based on RBF Adaptive control of neural networks and predictive control based on constrained particle swarm optimization are used to complete active pantograph control.
[0011] The beneficial effects of the present invention are:
[0012] 1. The present invention designs three controllers with different actuation positions according to the different installation positions of the controller, thereby reducing the fluctuation of the contact force between the bow and the catenary and further improving the current receiving quality of the bow and the catenary.
[0013] 2. The present invention only needs to obtain the current operating status information of the pantograph-catenary system and utilizes the control actuator to adjust and control the pantograph-catenary contact force, thus avoiding the deficiencies of existing optimization measures.
[0014] 3. The present invention takes into account the property of radial basis neural network to estimate disturbance, and designs RBF The adaptive control method of the neural network achieves compensation for model uncertainty by simultaneously controlling the actuation of the three mass blocks of the reduced mass model, thereby reducing the impact of disturbances and achieving the effect of reducing contact force fluctuations.
[0015] 4. Under the condition of actuator constraints or time lag, the system can still have better control stability and better trajectory tracking performance, which can effectively ensure the stability of pantograph operation, thereby achieving the purpose of improving current collection quality.
[0016] Preferably, the pantograph-contact network system simulation model in step S1 includes a pantograph reduced mass model, a multi-rigid body model and a contact network finite element model coupled by a penalty function; the pantograph reduced mass model includes three mass blocks.
[0017] The beneficial effects of the above preferred solution are:
[0018] Subdividing the pantograph-catenary system simulation model into the pantograph reduced mass model, multi-rigid body model and catenary finite element model will help to more accurately understand the operation mechanism and performance of the pantograph-catenary system and design more effective control strategies.
[0019] Preferably, the step S2 specifically comprises the following steps:
[0020] S21. Construct a collaborative dynamics model based on the pantograph mass model and the contact network finite element model in the pantograph-contact network system simulation model;
[0021] S22. The operation status of the pantograph-contact network system is simulated by a collaborative dynamics model to obtain simulation results;
[0022] S23. Design incremental type based on simulation results PID The controller is installed on the pantograph frame to provide control force.
[0023] As a preference, the incremental method in step S23 PID The expression is:
[0024]
[0025] in, To control the input increment, , For the controller Output at the moment, For the controller Output at the moment; For the controller Input error at the moment; For the controller Input error at the moment; For the controller Input error at the moment; is a positive proportional gain; is the integral gain; is the negative differential gain.
[0026] The beneficial effects of the above preferred solution are:
[0027] The control target of the controller is to reduce the standard deviation of the pantograph-catenary contact force fluctuation, that is, to reduce the difference between the real-time pantograph-catenary contact force and the target contact force. PID The control has the advantages of simple control principle, convenient parameter adjustment, easy implementation and wide applicability.
[0028] Preferably, the step S3 specifically comprises the following steps:
[0029] S31. Simplify the contact network finite element model to obtain a control-oriented pantograph-contact network system simulation model;
[0030] S32. Design of a control-oriented pantograph-catenary system simulation model PD Controller;
[0031] S33. PD Add to controller RBFThe neural network performs feed-forward compensation;
[0032] S34.Based on Lyapunov Theorem 2 Method, Design Lyapunov function;
[0033] S35.Based on Lyapunov Function Compute RBF The adaptive weight update law of the neural network is obtained based on RBF The adaptive controller of the neural network ensures the limited compensation of the control system.
[0034] As a preference, in step S32 PD The control law of the controller is:
[0035]
[0036] in, represents the control law; represents the mass parameter matrix of the pantograph parachute mass model, , represents the equivalent mass of the pantograph head, represents the equivalent mass of the pantograph upper frame, represents the equivalent mass of the pantograph lower frame; represents the damping parameter matrix, , represents the equivalent damping of the pantograph head, represents the equivalent damping of the pantograph upper frame, represents the equivalent damping of the pantograph lower frame; represents the stiffness parameter matrix, , represents the equivalent stiffness of the pantograph head, represents the equivalent stiffness of the pantograph upper frame, represents the equivalent stiffness of the pantograph lower frame, Indicates the contact network stiffness; represents the expected reference acceleration trajectory of the pantograph; represents the tracking error, , represents the expected reference displacement trajectory of the pantograph; represents a positive definite constant third-order constant matrix, , , Representation Matrix The adjustment parameters of Represents the first-order differential function of the tracking error; Indicates the current vertical displacement state of the pantograph; Indicates the vertical speed of the pantograph; represents the vehicle body disturbance and model parameter perturbation;
[0037] In step S34 Lyapunov The function is:
[0038]
[0039] in, express Lyapunov function; represents a symmetric positive definite matrix; express RBF Input of neural network; superscript Represents the transpose of a matrix; represents the optimal weight; Represents the optimal weight An estimated value of Represents the adjustment parameter, which is generally a positive constant;
[0040] In step S35 RBF The calculation formula of the adaptive weight update law of the neural network is:
[0041]
[0042] in, express RBF Adaptive weight update law for neural networks; represents a constant matrix; express RBF Activation functions for neural networks.
[0043] The beneficial effects of the above preferred solution are:
[0044] use RBF The neural network controller can effectively estimate model uncertainties and further improve the robustness and dynamic performance of the controller through disturbance compensation.
[0045] Preferably, the step S4 specifically comprises the following steps:
[0046] S41. Design a predictive controller with input constraints based on the control input constraints caused by the saturation of the actuator itself and the pantograph head lifting amount constraints in the pantograph-catenary system simulation model;
[0047] S42. According to the time lag problem of the actuator in the pantograph-catenary system simulation model, a predictive controller is designed that takes into account the time lag of the actuator;
[0048] S43. Based on the predictive controller with input constraints and the predictive controller considering the actuator time delay, the optimal control sequence is solved by using the particle swarm optimization algorithm to obtain the predictive controller based on constrained particle swarm optimization;
[0049] S44. Using the extended Kalman filter to obtain the status information of the pantograph in operation;
[0050] S45. According to the state information of the pantograph, the constraints and time lag problems existing in the actuator are controlled and compensated through a predictive controller based on constrained particle swarm optimization.
[0051] Preferably, the step S41 specifically includes the following steps:
[0052] S411. According to the control input constraint caused by the saturation of the actuator itself and the pantograph bow head lifting amount constraint in the pantograph-contact network system simulation model, an optimization equation containing input constraints is designed. The optimization equation containing input constraints is:
[0053]
[0054] in, represents the minimum value of the optimization equation; The state variables representing the extended state matrix; Indicates the input of the system at the last moment; Represents the incremental input of the system at the current moment; Represents the input of the system at the current moment; Indicates system input; subscript Indicates the minimum value of the parameter; subscript Indicates the maximum value of the parameter;
[0055] S412. Replace the optimization equation with input constraints The constraint function is transformed into The functional relationship of is obtained to obtain the transformed optimization equation;
[0056] S413. The converted optimization equation is converted into an optimization equation in a standard quadratic form, wherein the optimization equation in the standard quadratic form is:
[0057]
[0058] in, represents the optimization equation; Indicates a new control sequence; ; ; represents the error weight matrix; represents the control input weight matrix; The error matrix representing the current contact force of the pantograph-catenary system and the reference contact force; represents the state matrix; superscript Represents the transpose of a matrix; represents the incremental input matrix; represents the constraint coefficient; represents the constraint matrix;
[0059] S414. Solve the optimization equation in standard quadratic form to obtain the control input of the pantograph-catenary system , that is, a predictive controller with input constraints is obtained.
[0060] The beneficial effects of the above preferred solution are:
[0061] Under the condition of actuator constraints or time lag, the system can still have better handling stability and better trajectory tracking performance, which can effectively ensure the stability of pantograph operation, thereby achieving the purpose of improving the current collection quality.
[0062] Preferably, the particles of the particle swarm optimization algorithm in step S43 are set to predict the input increment, and the dimension of the particle swarm is D Set to predictive control horizon Nc , the fitness function of the particle swarm is set to the equation to be optimized.
[0063] Preferably, step S45 specifically includes the following steps:
[0064] S451. Initialize the control parameters;
[0065] S452. k At this moment, the error matrix is calculated based on the initialized control parameters, system state information, output information and constraints. , And the constraint matrix , and k The control sequence at -1 is k Initialization value of particle swarm at time;
[0066] S453. Optimization using particle swarm optimization algorithm, based on the error matrix , and the constraint matrix as well as k The initialization value of the particle swarm at time t is solved to obtain a new control sequence ,Will The first component of ;
[0067] S454. Increase the optimal control increment Input system for control compensation;
[0068] S455. Determine whether the pantograph-contact network system stability is optimal based on the pantograph status information. If so, the system control compensation is completed. If not, set the count value Add 1 and return to step S451.
[0069] The beneficial effects of the above preferred solution are:
[0070] By introducing the particle swarm optimization algorithm into the predictive control model and using the particle swarm optimization algorithm to solve the minimum value of the fitness function, the optimal control sequence can be solved to effectively ensure the stability of the pantograph operation under the conditions of actuator constraints or time lags. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] Figure 1 Shown is a flow chart of a pantograph active control method considering controlling different components. DETAILED DESCRIPTION
[0072] Now, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the accompanying drawings are only exemplary and are intended to explain the principles and spirit of the present invention, rather than to limit the scope of the present invention.
[0073] like Figure 1 As shown, a pantograph active control method considering controlling different components comprises the following steps:
[0074] S1. Establish a simulation model of the pantograph-catenary system;
[0075] S2. According to the pantograph-contact network system simulation model, the pantograph head and frame are taken as the control objects, and the proportional-integral-differential (PID) PID ) controller;
[0076] S3. According to the pantograph-contact network system simulation model, the pantograph head and the upper frame are simultaneously taken as control objects, and the radial basis function ( RBF ) Adaptive controllers based on neural networks;
[0077] S4. According to the pantograph-catenary system simulation model, the pantograph chassis is taken as the control object, and a predictive controller based on constrained particle swarm optimization is designed;
[0078] S5. Based on PID Controller based on RBF Adaptive control of neural networks and predictive control based on constrained particle swarm optimization are used to complete active pantograph control.
[0079] In this embodiment, the pantograph-contact network system simulation model in step S1 includes a pantograph reduced mass model, a multi-rigid body model and a contact network finite element model coupled by a penalty function; the pantograph reduced mass model includes three mass blocks.
[0080] In this embodiment, step S2 specifically includes the following steps:
[0081] S21. According to the pantograph mass model and the contact network finite element model in the pantograph-contact network system simulation model, a Simulink Collaborative dynamics model with dynamics 3D simulation software;
[0082] S22. The operation status of the pantograph-contact network system is simulated by a collaborative dynamics model to obtain simulation results;
[0083] S23. Design incremental type based on simulation results PID The controller is installed on the pantograph frame to provide control force.
[0084] In this embodiment, the incremental PID The expression is:
[0085]
[0086] in, To control the input increment, , For the controller Output at the moment, For the controller Output at the moment; For the controller Input error at the moment; For the controller Input error at the moment; For the controller Input error at the moment; is a positive proportional gain; is the integral gain; is the negative differential gain.
[0087] In this embodiment, step S3 specifically includes the following steps:
[0088] S31. Simplify the contact network finite element model to obtain a control-oriented pantograph-contact network system simulation model;
[0089] S32. Design of proportional-differential ( PD ) Controller;
[0090] S33. PD Add to controller RBF The neural network performs feed-forward compensation;
[0091] S34. Based on Lyapunov ( Lyapunov ) Theorem 2 Method, Design Lyapunov function;
[0092] S35.Based on Lyapunov Function Compute RBF The adaptive weight update law of the neural network is obtained based on RBF The adaptive controller of the neural network ensures the limited compensation of the control system.
[0093] In this embodiment, the step S32 PD The control law of the controller is:
[0094]
[0095] in, represents the control law; represents the mass parameter matrix of the pantograph mass block model, , represents the equivalent mass of the pantograph head, represents the equivalent mass of the pantograph upper frame, represents the equivalent mass of the pantograph lower frame; represents the damping parameter matrix, , represents the equivalent damping of the pantograph head, represents the equivalent damping of the pantograph upper frame, represents the equivalent damping of the pantograph lower frame; represents the stiffness parameter matrix, , represents the equivalent stiffness of the pantograph head, represents the equivalent stiffness of the pantograph upper frame, represents the equivalent stiffness of the pantograph lower frame, Indicates the contact network stiffness; represents the expected reference acceleration trajectory of the pantograph; represents the tracking error, , represents the expected reference displacement trajectory of the pantograph; represents a positive definite constant third-order constant matrix, , , Representation Matrix The adjustment parameters of Represents the first-order differential function of the tracking error; Indicates the current vertical displacement state of the pantograph; Indicates the vertical speed of the pantograph; represents the vehicle body disturbance and model parameter perturbation;
[0096] In step S34 Lyapunov The function is:
[0097]
[0098] in, express Lyapunov function; represents a symmetric positive definite matrix; express RBF Input of neural network; superscript Represents the transpose of a matrix; represents the optimal weight; Represents the optimal weight An estimated value of represents the adjustment parameter;
[0099] In step S35 RBF The calculation formula of the adaptive weight update law of the neural network is:
[0100]
[0101] in, express RBF Adaptive weight update law for neural networks; represents a constant matrix; express RBF Activation functions for neural networks.
[0102] In this embodiment, step S4 specifically includes the following steps:
[0103] S41. Design a predictive controller with input constraints based on the control input constraints caused by the saturation of the actuator itself and the pantograph head lifting amount constraints in the pantograph-catenary system simulation model;
[0104] S42. According to the time lag problem of the actuator in the pantograph-catenary system simulation model, a predictive controller is designed that takes into account the time lag of the actuator;
[0105] S43. Based on the predictive controller with input constraints and the predictive controller considering the actuator time delay, the optimal control sequence is solved by using the particle swarm optimization algorithm to obtain the predictive controller based on constrained particle swarm optimization;
[0106] S44. Using the extended Kalman filter to obtain the status information of the pantograph in operation;
[0107] S45. According to the state information of the pantograph, the constraints and time lag problems existing in the actuator are controlled and compensated through a predictive controller based on constrained particle swarm optimization.
[0108] In this embodiment, step S41 specifically includes the following steps:
[0109] S411. According to the control input constraint caused by the saturation of the actuator itself and the pantograph bow head lifting amount constraint in the pantograph-contact network system simulation model, an optimization equation containing input constraints is designed. The optimization equation containing input constraints is:
[0110]
[0111] in, represents the minimum value of the optimization equation; The state variables representing the extended state matrix; Indicates the input of the system at the last moment; Represents the incremental input of the system at the current moment; Represents the input of the system at the current moment; Indicates system input; subscript Indicates the minimum value of the parameter; subscript Indicates the maximum value of the parameter;
[0112] S412. Replace the optimization equation with input constraints The constraint function is transformed into The functional relationship of is obtained to obtain the transformed optimization equation;
[0113] S413. The converted optimization equation is converted into an optimization equation in a standard quadratic form, wherein the optimization equation in the standard quadratic form is:
[0114]
[0115] in, represents the optimization equation; Indicates a new control sequence; ; ; represents the error weight matrix; represents the control input weight matrix; The error matrix representing the current contact force of the pantograph-catenary system and the reference contact force; represents the state matrix; superscript Represents the transpose of a matrix; represents the incremental input matrix; represents the constraint coefficient; represents the constraint matrix;
[0116] S414. Solve the optimization equation in standard quadratic form to obtain the control input of the pantograph-catenary system , that is, a predictive controller with input constraints is obtained.
[0117] In this embodiment, the particles of the particle swarm optimization algorithm in step S43 are set to predict the input increment, and the dimension of the particle swarm is D Set to predictive control horizon Nc , the fitness function of the particle swarm is set as the equation to be optimized. Then the process of the particle swarm optimization algorithm to solve the minimum value of the fitness function is to solve the predictive control optimization equation and find the optimal control sequence.
[0118] In this embodiment, step S45 specifically includes the following steps:
[0119] S451. Initialize the control parameters;
[0120] S452. k At this moment, the error matrix is calculated based on the initialized control parameters, system state information, output information and constraints. , And the constraint matrix , and k The control sequence at -1 is k Initialization value of particle swarm at time;
[0121] S453. Optimization using particle swarm optimization algorithm, based on the error matrix , and the constraint matrix as well as k The initialization value of the particle swarm at time t is solved to obtain a new control sequence ,Will The first component of ;
[0122] S454. Increase the optimal control increment Input system for control compensation;
[0123] S455. Determine whether the pantograph-contact network system stability is optimal based on the pantograph status information. If so, the system control compensation is completed. If not, set the count value Add 1 and return to step S451.
[0124] The effect of the present invention is verified below:
[0125] 1. Based on PID Verification of controller performance
[0126] Using the pantograph-catenary system simulation model, the simulation results of the pantograph-catenary system were carried out at two speeds: 200km / h and 300km / h. PID The control performance of the controller is analyzed.
[0127] Based on the reference contact force F The calculation formula is: F =0.00097 v 2 +70, set the reference contact force at 200km / h to 108.8N, and the controller parameters are k p =0.9, k i =0, k d =-0.008. The simulation calculation results are shown in Table 1. It can be clearly seen that when the train speed is 200km / h, the controller can effectively reduce the fluctuation of the contact force while keeping the average value of the contact force basically unchanged (the average value of the contact force increases by 0.74%), and the standard deviation of the contact force decreases by 19.06%. At the same time, the maximum value and statistical maximum value of the contact force decrease, while the minimum value and statistical minimum value increase. The reduction in the vertical amplitude fluctuation of the pantograph head can effectively improve the current collection quality of the pantograph network and improve the safety of train operation.
[0128] Table 1 Comparison of simulation results before and after 200km / h control
[0129]
[0130] According to the formula F =0.00097 v 2 +70 sets the reference contact force at 300 km / h to 157.3N. PID The parameters of the controller are k p =1.3, k i =0, k d =-0.01. The simulation calculation results are shown in Table 2. It can be clearly seen that when the train speed is 300km / h, by adjusting the controller parameters, the controller can still effectively reduce the fluctuation of the contact force while keeping the average value of the contact force basically unchanged (the average value of the contact force increases by 1.44%). The standard deviation of the contact force is reduced by 13.01%. Although the performance of the controller is weakened compared to 200km / h, it still meets the requirements of the maximum value and statistical maximum value of the contact force, while the minimum value and statistical minimum value are improved. In summary, the proposed control strategy can effectively reduce the standard deviation of the contact force at different speed levels, which verifies the proposed PID Effectiveness of control strategies.
[0131] Table 2 Comparison of simulation results before and after 300km / h control
[0132]
[0133] 2. Based on RBF Analysis of Control Effect of Adaptive Controller Based on Neural Network
[0134] The train speed is set to 300km / h, and the pantograph model is DSA 380 pantograph, contact network parameters, suspension string spacing L d =50m; average stiffness k 0=1925N / m; fitting coefficient of contact network stiffness change: α 1=0.0755, α 2 = -0.0735, α 3=-0.1459, α 4=-0.0575, α 5=0.0699; neural network Gaussian basis function width b =7.07; Number of neuron nodes n is 128; according to the state information of the bow-net model, the center vector of the Gaussian basis function is designed c =[ h p I , h v I , h p I , h v I , h p I , h v I ] T ⊗ c 1, where I represents the identity matrix, h p , h v is the regulating factor, h p =0.1, h v =0.5; c1=[2 / n (- n / 2,...,-1,0,1,..., n / 2-1), where nIndicates the number of neural network nodes; PD Control parameters α =10; initial weight of neural network w =0, Q =100* eye (6) γ =0.5. Optimal control LQR In , the state weight matrix is set to diag [100,1,100,1,100,1], the control weight coefficient is 10 -6 eye (6).
[0135] In the absence of disturbance, the pantograph-catenary reference contact force is set to 70N, that is:
[0136]
[0137] in, is the train running speed, is the train running time, is the span, L d represents the distance between the suspension strings, k 0 means average stiffness, α 1, α 2, α 3. α 4. α 5 is the fitting coefficient of contact network stiffness change, is the contact network stiffness.
[0138] Under the condition of no disturbance, the linear quadratic optimal control ( LQR The standard deviation of contact force was reduced by 12.79% compared with that before control. PD The control can complete the tracking control of the reference contact force with almost no overshoot. PD The maximum lifting displacement of the bow head after control is 0.048m. LQR The maximum lifting displacement of the controlled bow head is 0.036m, and the lifting displacement of the bow head without control is 0.043m. PD Although the control effect on contact force is better than LQR Optimal control, but will slightly increase the fluctuation of the vertical vibration displacement of the bow head. As for the vertical displacement of the upper frame and the vertical displacement of the lower frame, PD Control can effectively reduce its fluctuations.
[0139] Consider the case where there is a random external disturbance in the pantograph-catenary system. d Set to d = r + 2 *sint ,in r represents disturbance, r =[ r 1, r 2, r 3] The value range is r 1=[0,2], r 2=[0,5], r 3=[0,10] is a random number. RBF Neural networks can effectively approximate external random disturbances, thereby effectively compensating for the impact of external random disturbances. PD Control, due to RBF Neural networks can effectively estimate the disturbances to the system and compensate for them. RBF-PD It has stronger robustness. When there are random disturbances, it shows better dynamic characteristics and anti-interference ability.
[0140] Consider the case where the model is uncertain, that is, when there is both parameter mismatch and external disturbance. RBF Neural network estimation performance and controller performance. Even in the presence of parameter mismatch and external disturbance, the designed RBF Neural networks can still effectively approximate external random disturbances and compensate for them. Consider using R BF The neural network compensates for the uncertainty of the model by fitting, and its contact force and the trajectory tracking effect of the vertical displacement of the bow head are significantly better than those without considering the compensation. PD control.
[0141] 3. Analysis of control effect of predictive controller based on constrained particle swarm optimization
[0142] Pantograph model adopts DSA 380 pantograph, contact network parameters, suspension string spacing L d =50m; average stiffness k 0=1925N / m; fitting coefficient of contact network stiffness change: α 1=0.0755, α 2 = -0.0735, α 3=-0.1459, α 4=-0.0575, α 5=0.0699; Considering the external interference of vehicle body vibration, verify the effectiveness and robustness of the control strategy at different operating speeds. Use white noise with a power of 0.005 to add to the pantograph-network system to simulate the interference caused by vehicle body vibration on the running pantograph. Assume that the measurement noise v For 10 -3 diag(1,1,1), obtain state information through displacement sensor. Set the initial value of Kalman filter algorithm: , , , , v =250km / h. By estimating the measured vertical displacement information (including noise interference), the state estimates of the three mass blocks (displacement estimate information and velocity estimate information) are obtained. The root mean square error of the estimated value of the pantograph state is shown in Table 3. From the root mean square error analysis of the estimated values of the parameters in the table, it can be seen that using the extended Kalman filter as an observer is feasible and reliable.
[0143] Table 3 Root mean square error of pantograph state estimation
[0144]
[0145] Consider the unconstrained model predictive control algorithm ( MPC ), as can be seen from the cost function, the selection of weights not only affects the performance of the controller, but is also related to the energy consumption of the control input. Therefore, it is necessary to adjust the weight coefficient to obtain a balance between the above performances while ensuring that the fluctuation amplitude of the contact force is reduced as much as possible. The analysis is carried out under two speed conditions of 250km / h and 350km / h, and the reference contact force is set to the dynamic contact force of the nominal nonlinear bow-net model, denoted as Y ref , static contact force of bow-catcher F 0=70N, sampling time T =0.01s; control parameters Np =10, Nc =4, Q = eye ( Np ), R =10 -4 eye ( Nu ). When considering the interference caused by the vibration of the car body, the standard deviation of the bow-net contact force increased by 36.32% and 39.36% with the increase of speed level; the standard deviation of the contact force decreased by 35.23% and 33.56%, respectively, and the maximum value of the contact force decreased and the minimum value increased after control. Although the performance of the controller is gradually weakening with the increase of the train speed, the controller can still reduce the fluctuation of the contact force. After the train runs stably, the maximum lifting displacement of the bow head under control is 0.047m, while the lifting displacement of the bow head when no control is applied reaches 0.08m. The active control force range of the controller at both speed levels is between -118 and 63N, and the active control force jitter is greater at 350km / h. Therefore, with the increase of speed level, the requirements for the actuator are also higher.
[0146] Consider the constraints MPC The algorithm can limit the output range of the actuator by constraining the active control input force, which is more in line with the application requirements in actual engineering. The reference contact force is set to the dynamic contact force of the nominal nonlinear pantograph model, denoted as Y ref , static contact force of bow-catcher F 0=70N, sampling time T =0.01s; control parameters Np =10, Nc =4, Q =eye( Np ), R =10 -4 eye( Nu ); Particle swarm parameter settings w s =0.9, w e =0.4, c 1s =2.5, c 2s =0.5, c 1e =0.5, c 2e =2.5, the number of particle swarms is 80, the maximum number of iterations N =250. After multiple simulation experiments, the range of active control input force is set to -60~60N, and the range of single increment of control input force is -50~50N. MPC Less restrictive control MPC The performance remains basically unchanged. Under the same external disturbance, considering the constraints MPC Under the control, the standard deviation of the pantograph-catenary contact force at the two speed levels was reduced by 35.75% and 33.84%, respectively. MPC Control its active control input compared to the unconstrained MPC The input force is constrained and limited, which effectively reduces the actuator's actuation range. This can effectively reduce the actuator's performance requirements and energy consumption while achieving better performance indicators. MPC From the control input force, it can be found that the required control input force waveform at 350 km / h has a more severe oscillation and jitter, which further shows that with the increase of speed, the actuator needs more energy to meet the performance indicators.
[0147] Consider the impact of input lag on control performance. Set control parameters Np =10, Nc=4, Q =eye( Np ), R =10 -4 eye( Nu ); Particle swarm parameter settings w s =0.9, w e =0.4, c 1s =2.5, c 2s =0.5, c 1e =0.5, c 2e =2.5, the number of particle swarms is 80, the maximum number of iterations N =250. Consider the impact on control performance when the input time lag is 1ms and 10ms. At 250km / h, the range of active control input force is set to -60~60N, and the range of single increment of control input force is -50~50N; at 350km / h, the range of active control input force is set to -100~100N, and the range of single increment of control input force is -80~80N. Analysis shows that when the time lag is 1ms, MPC The control performance is basically not affected, and the bow-net contact force and the vertical displacement of the bow head can still be effectively controlled. When the time lag is 1ms, the input force constraint is effectively limited to -60~60N, but the control input force required for control has a significant oscillation amplitude, and the energy required is larger than when there is no time lag, and the energy consumption is higher. When the time lag is 10ms, that is, the time lag is one control sampling step, MPC The control performance is greatly reduced. The bow-net contact force and bow head vertical displacement curves fluctuate more violently at two speed levels. The standard deviation and maximum value of the contact force increase, while the minimum value decreases, and even produces certain negative effects. When the time lag is 10ms, although the input force is effectively limited to -100~100N, the control input force required for control oscillates more violently, requires more energy, and consumes more energy. It can be seen that the time lag will inhibit the control performance, and the greater the input time lag, the more unfavorable it is for the active control design of the pantograph.
[0148] Through the above analysis, the following is given MPC Delay compensation controller and MPC The performance comparison of the controller when the control input delay is 10ms. MPC The controller can effectively reduce the impact of input time delay, reduce the standard deviation of the pantograph-catenary contact force and the vertical displacement of the bow head, and improve the current-collecting quality of the pantograph-catenary system. MPCAfter the control, the standard deviation of the contact force is reduced by 23.28% and 25.72% respectively compared with the MPC control without time-delay compensation. MPC The controller consumes less energy. At the same time, the input force is constrained. After the controller controls, the input force fluctuation range is limited to -60~60N and -100~100N respectively.
[0149] In summary, MPC Time-delay compensation control can effectively compensate for the weakening of control performance caused by the increase of time-delay.
[0150] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.
Claims
1. A pantograph active control method considering the control of different components, It is characterized in that The following steps are involved: S1. Establish a simulation model of the pantograph-catenary system; The pantograph-contact network system simulation model includes a pantograph reduced mass model, a multi-rigid body model and a contact network finite element model coupled by a penalty function; the pantograph reduced mass model includes three mass blocks; S2. According to the pantograph-contact network system simulation model, the pantograph head and frame are taken as control objects, and the design is based on PID The controller specifically includes the following steps: S21. Construct a collaborative dynamics model based on the pantograph mass model and the contact network finite element model in the pantograph-contact network system simulation model; S22. The operation status of the pantograph-contact network system is simulated by a collaborative dynamics model to obtain simulation results; S23. Design incremental PID The controller is installed on the upper frame of the pantograph to provide control force; S3. According to the pantograph-contact network system simulation model, the pantograph head and the upper frame are simultaneously taken as control objects, and the design is based on RBF The adaptive controller of the neural network specifically includes the following steps: S31. Simplify the contact network finite element model to obtain a control-oriented pantograph-contact network system simulation model; S32. Design of a control-oriented pantograph-catenary system simulation model PD Controller; S33. PD Add to controller RBF The neural network performs feed-forward compensation; S34.Based on Lyapunov Theorem 2 Method, Design Lyapunov function; S35.Based on Lyapunov Function Compute RBF The adaptive weight update law of the neural network is obtained based on RBF Adaptive controller of neural network to ensure limited compensation of control system; S4. According to the pantograph-contact network system simulation model, the pantograph chassis is taken as the control object, and a predictive controller based on constrained particle swarm optimization is designed, which specifically includes the following steps: S41. Design a predictive controller with input constraints based on the control input constraints caused by the saturation of the actuator itself and the pantograph head lifting amount constraints in the pantograph-catenary system simulation model; S42. According to the time lag problem of the actuator in the pantograph-catenary system simulation model, a predictive controller is designed that takes into account the time lag of the actuator; S43. Based on the predictive controller with input constraints and the predictive controller considering the actuator time delay, the optimal control sequence is solved using the particle swarm optimization algorithm to obtain a predictive controller based on constrained particle swarm optimization. The particles of the particle swarm optimization algorithm are set to predict the input increment, and the dimension of the particle swarm is D Set to predictive control horizon Nc , the fitness function of the particle swarm is set to the equation to be optimized; S44. Using the extended Kalman filter to obtain the status information of the pantograph in operation; S45. According to the state information of the pantograph, the constraints and time delay problems of the actuator are controlled and compensated by a predictive controller based on constrained particle swarm optimization; The step S41 specifically includes the following steps: S411. According to the control input constraint caused by the saturation of the actuator itself and the pantograph bow head lifting amount constraint in the pantograph-contact network system simulation model, an optimization equation containing input constraints is designed. The optimization equation containing input constraints is: in, represents the minimum value of the optimization equation; The state variables representing the extended state matrix; Indicates the input of the system at the last moment; Represents the incremental input of the system at the current moment; Represents the input of the system at the current moment; Indicates system input; subscript Indicates the minimum value of the parameter; subscript Indicates the maximum value of the parameter; S412. Replace the optimization equation with input constraints The constraint function is transformed into The functional relationship of is obtained to obtain the transformed optimization equation; S413. The converted optimization equation is converted into an optimization equation in a standard quadratic form, wherein the optimization equation in the standard quadratic form is: in, represents the optimization equation; Indicates a new control sequence; ; ; represents the error weight matrix; represents the control input weight matrix; The error matrix representing the current contact force of the pantograph-catenary system and the reference contact force; represents the state matrix; superscript Represents the transpose of a matrix; represents the incremental input matrix; represents the constraint coefficient; represents the constraint matrix; S414. Solve the optimization equation in standard quadratic form to obtain the control input of the pantograph-catenary system , that is, a predictive controller with input constraints is obtained; The step S45 specifically includes the following steps: S451. Initialize the control parameters; S452. k At this moment, the error matrix is calculated based on the initialized control parameters, system state information, output information and constraints. , And the constraint matrix , and k The control sequence at -1 is k Initialization value of particle swarm at time; S453. Optimization using particle swarm optimization algorithm, based on the error matrix , and the constraint matrix as well as k The initialization value of the particle swarm at time t is solved to obtain a new control sequence ,Will The first component of ; S454. Increase the optimal control increment Input system for control compensation; S455. Determine whether the pantograph-contact network system stability is optimal based on the pantograph status information. If so, the system control compensation is completed. If not, set the count value Add 1 and return to step S451; S5. Based on PID Controller based on RBF Neural network adaptive controller and constrained particle swarm optimization-based predictive controller to achieve active pantograph control; The incremental method in step S23 PID The expression is: in, To control the input increment, , For the controller The output of the moment, For the controller Output at the moment; For the controller Input error at the moment; For the controller Input error at the moment; For the controller Input error at the moment; is a positive proportional gain; is the integral gain; is the negative differential gain; In step S32 PD The control law of the controller is: in, represents the control law; represents the mass parameter matrix of the pantograph three-mass model, , represents the equivalent mass of the pantograph head, represents the equivalent mass of the pantograph upper frame, represents the equivalent mass of the pantograph lower frame; represents the damping parameter matrix, , represents the equivalent damping of the pantograph head, represents the equivalent damping of the pantograph upper frame, represents the equivalent damping of the pantograph lower frame; represents the stiffness parameter matrix, , represents the equivalent stiffness of the pantograph head, represents the equivalent stiffness of the pantograph upper frame, represents the equivalent stiffness of the pantograph lower frame, Indicates the contact network stiffness; represents the expected reference acceleration trajectory of the pantograph; represents the tracking error, , represents the expected reference displacement trajectory of the pantograph; represents a positive definite constant third-order constant matrix, , , Representation Matrix The adjustment parameters of Represents the first-order differential function of the tracking error; Indicates the current vertical displacement state of the pantograph; Indicates the vertical speed of the pantograph; represents the vehicle body disturbance and model parameter perturbation; In step S34 Lyapunov The function is: in, express Lyapunov function; represents a symmetric positive definite matrix; express RBF Input of neural network; superscript Represents the transpose of a matrix; represents the optimal weight; Represents the optimal weight An estimated value of represents the adjustment parameter; In step S35 RBF The calculation formula of the adaptive weight update law of the neural network is: in, express RBF Adaptive weight update law for neural networks; represents a constant matrix; express RBF Activation functions for neural networks.
Citation Information
Patent Citations
Experimental pantograph and catenary system dynamic pressure precise loading device and control method
CN105388760A
Electric locomotive pantograph and arc detection method based on deep network and graded testing
CN108288055A
Contact network prior information-based active control method of pantograph neural network sliding-mode variable structure
CN110850716A
Self-adaptive control method and device of pantograph-catenary system based on reference model, and medium
CN117572772A
Power system of hybrid truck and hybrid truck
CN211364273U
