Active control method for straddle-type monorail pantograph based on prediction of pantograph-catenary contact force

Abstract

The present application relates to the straddle-type monorail pantograph active control method based on the prediction of the pantograph-catenary contact force, belongs to the pantograph active control field, and comprises the following steps: S1: obtaining the state characteristic data and the actual pantograph-catenary contact force data of the pantograph-catenary system in the running process of the straddle-type monorail train; S2: constructing the intelligent model of the straddle-type monorail pantograph-catenary system based on the ''mechanism-data'' double drive; S3: extracting the time sequence characteristics of the pantograph-catenary contact force, inputting into the IGWO-GRU neural network, and constructing the prediction model of the straddle-type monorail pantograph-catenary contact force; S4: constructing the nonlinear model predictive control framework of the straddle-type monorail pantograph; S5: using the sequential quadratic programming method (SQP) to perform online solution on the rolling optimization objective function; the first control increment of the optimal control increment sequence is used on the pantograph actuator, and the closed-loop correction mechanism is constructed. The present application reduces the pantograph-catenary contact force fluctuation degree, improves the current collection quality of the train, and improves the service life of the pantograph slide.

Description

Technical Field

[0001] This invention belongs to the field of active pantograph control and relates to an active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction. Background Technology

[0002] As an important system in modern urban rail transit, straddle-type monorails typically use a rigid overhead contact line arranged on the side for power supply. Compared to the flexible catenary suspension of high-speed railways, the rigid overhead contact line has extremely high overall equivalent stiffness and lacks flexible buffering. When the train is running at high speed, curves, or encounters hard points in the rigid overhead contact line joints or track irregularities, the coupled vibration of the car body and the track beam will cause severe nonlinear fluctuations in the pantograph-catenary contact force. Excessive contact force will exacerbate abnormal mechanical wear of the carbon sliding contact plate, while insufficient contact force can easily lead to pantograph-catenary derailment and arcing, seriously threatening the safety of train traction power supply.

[0003] To suppress fluctuations in the contact force of straddle-type monorail pantographs, several active control methods for straddle-type monorail pantographs have been proposed in the prior art. For example, Chinese patent document CN202111356138.1, entitled "An Invention Patent for a Straddle-Type Intelligent Monorail Pantograph," discloses a scheme that predicts the contact force at the next moment by embedding a standard gated recurrent unit (GRU) neural network in the pantograph controller. It then directly uses the difference between the expected and predicted contact force values ​​as the desired control force, sending the expected value to the magnetorheological damper to perform active control.

[0004] The aforementioned authorized active control scheme for straddle-type monorail pantographs improves the current collection quality of trains to some extent. However, when dealing with the continuous speed increase of straddle-type monorail trains and the more severe and complex high-frequency disturbances to the rigid contact network (such as continuous impacts at joint hard points), the scheme reveals the following defects:

[0005] 1) The prediction model lacks global optimization and suffers from prediction lag and local optima:

[0006] The above scheme only uses a basic GRU neural network. Due to the highly nonlinear and time-varying nature of the pantograph-catenary contact force data, if the hyperparameters of the basic GRU network (such as the number of hidden layer neurons and the learning rate) are set based on manual experience, they are prone to getting trapped in local optima. When encountering a step-like abrupt shock caused by a hard point in the rigid catenary, the basic GRU model cannot promptly cut off the transmission of historical steady-state inertia, resulting in severe "amplitude decay" and "phase lag" in the prediction curve, making accurate advance prediction impossible.

[0007] 2) The control logic is simple, which easily leads to high-frequency chattering and mechanical fatigue in the actuators:

[0008] The control algorithm described above simply calculates the control force by subtracting the "expected value" from the "predicted value" using arithmetic. This open-loop calculation method completely ignores the future dynamic evolution of the system (lacking a rolling optimization mechanism). More critically, when faced with high-frequency fluctuating contact forces, this method directly outputs an absolute control force signal with extremely large amplitude and frequency. This forces the underlying actuators to perform high-frequency forward and reverse switching (i.e., high-frequency chatter). Prolonged high-frequency chatter not only fails to effectively smooth out fluctuations but also induces destructive resonance in the pantograph frame, severely shortening the lifespan of the actuators.

[0009] 3) Lack of underlying algorithmic constraints on system physical boundaries and security requirements:

[0010] The above scheme, when calculating the control force, failed to take into account the physical constraints at the algorithm level, such as the maximum output thrust of the pantograph actuator, the response rate limit, and the upper and lower limits of the safe contact force to prevent the overhead contact line from breaking. Under extreme and severe operating conditions, it is very easy to output commands that exceed the actuator's execution capability, leading to control system failure. Summary of the Invention

[0011] In view of this, the purpose of this invention is to reduce the problem of severe fluctuations in pantograph-catenary contact force caused by rigid contact wire joints, hard points and track irregularities during straddle-type monorail operation, thereby reducing pantograph-catenary disconnection, improving the current collection quality of the vehicle, and extending the service life of the pantograph. This invention provides an active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction.

[0012] To achieve the above objectives, the present invention provides the following technical solution:

[0013] An active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction includes the following steps:

[0014] S1: Obtain the pantograph-catenary system status characteristic data and actual pantograph-catenary contact force data during the operation of straddle-type monorail trains;

[0015] S2: Construct an intelligent model of a straddle-type monorail pantograph-catenary system driven by both "mechanism and data" and use the PSO-XGBoost algorithm to accurately compensate for the nonlinear residuals of the pantograph-catenary binary quality model.

[0016] S3: Extract the timing features of the pantograph-catenary contact force and input them into the IGWO-GRU neural network to construct a straddle-type single-track pantograph-catenary contact force prediction model, so as to achieve accurate prediction of the pantograph-catenary contact force in the next control cycle.

[0017] S4: Construct a nonlinear model predictive control framework for straddle-type monorail pantographs, using the IGWO-GRU straddle-type monorail pantograph-catenary contact force prediction model as a nonlinear predictive operator, and combining the reference pantograph-catenary contact force with the actuator control increment to construct a rolling optimization objective function;

[0018] S5: Under the physical constraints of the straddle-type monorail pantograph-catenary system, the rolling optimization objective function is solved online using the Sequential Quadratic Programming (SQP) method to obtain the optimal control increment sequence; the first control increment of the optimal control increment sequence is extracted and applied to the pantograph actuator, and a closed-loop correction mechanism is constructed.

[0019] Furthermore, step S1 specifically includes:

[0020] Data on pantograph acceleration, displacement, temperature, and grating sensor wavelength were collected during actual vehicle testing on Chongqing Rail Transit Line 3. The OSC7100 Bragg grating accelerometer from Beijing Tongwei Technology, with a sensitivity of 0.17g, was selected. Its principle is to calculate acceleration by detecting changes in the center wavelength. Parameters were set as follows: wavelength range 1529–1564 nm, wavelength resolution 1 pm, and scanning frequency 1 kHz. The optical signal was then processed by a demodulator and converted into an acceleration signal. Finally, the pantograph-catenary contact force was calculated based on the measured acceleration data and the technical parameters of the monorail pantograph.

[0021] Furthermore, step S2 specifically includes:

[0022] S21: Construct a binary mass model for a straddle-type monorail pantograph-catenary system;

[0023] S22: Construct a nonlinear residual compensation model for contact force of straddle-type monorail pantograph-catenary system based on XGBoost;

[0024] S23: The Particle Swarm Optimization (PSO) algorithm is used to adaptively optimize hyperparameters such as the learning rate, maximum tree depth, and regularization coefficient of XGBoost.

[0025] Furthermore, step S21 specifically includes: simplifying the pantograph and the contact wire into a whole coupled together by an elastic element.

[0026] Treating the rigid contact wire as a time-varying stiffness system, the formula is as follows:

[0027]

[0028] Where: The average stiffness coefficient within the span, i.e. ;in , These are the maximum and minimum stiffness coefficients of the overhead contact system within one span. The stiffness variation coefficient of the overhead contact line is... ; The locomotive's operating speed; The contact wire span; This refers to the runtime.

[0029] The dynamic equation for a straddle-type monorail is:

[0030]

[0031] In the formula: The equivalent stiffness of the overhead contact line; , , The equivalent mass, equivalent damping, and vertical displacement of the bow head; , , The equivalent mass, equivalent damping, and vertical displacement of the frame; , The equivalent stiffness of the bow head support rubber and the lifting spring; This is the static lifting force of the pantograph; This is a vehicle body disturbance signal; This is an excitation signal for a rigid overhead contact line.

[0032] Transform the dynamic equations into state equations;

[0033] Let the state vector be The state-space equations are as follows:

[0034]

[0035] The coefficient matrix in the formula is as follows:

[0036]

[0037]

[0038]

[0039]

[0040] The formula for calculating the contact force between the pantograph and the catenary is as follows:

[0041]

[0042] Furthermore, step S22 specifically includes: using XGBoost's powerful nonlinear mapping capability to compensate for the high-frequency vibration characteristics lost due to the linear simplification truncation of the binary mass model, and fitting the nonlinear transient deviations caused by the irregularities and hard point impacts of the rigid contact network.

[0043] Nonlinear residual of pantograph-catenary contact force Defined as:

[0044]

[0045] In the formula: The measured pantograph-catenary contact force was collected by S1. This is the baseline pantograph-catenary contact force output by the binary mass model.

[0046] Furthermore, step S23 specifically includes: adaptively optimizing hyperparameters such as the learning rate, maximum tree depth, and regularization coefficient of XGBoost using a particle swarm optimization (PSO) algorithm. The reconstructed dynamic pantograph-catenary contact force... for:

[0047]

[0048] Furthermore, step S3 specifically includes:

[0049] S31: The reset and update gate mechanisms of the GRU are used to perform nonlinear fusion screening of the stiffness mutation of the rigid contact wire and the historical characteristics of the contact force of the pantograph-catenary system in straddle-type monorail pantograph-catenary system.

[0050] S32: The improved Grey Wolf Optimization (IGWO) algorithm is used to optimize the hyperparameters of GRU: Logistic chaotic mapping is introduced to generate a uniform initial population; a nonlinear dynamic convergence factor is introduced to dynamically adjust the step size of global exploration and local development.

[0051] The optimized output equation for pantograph-catenary contact force prediction by IGWO is:

[0052]

[0053] In the formula, In the current Always looking towards the future Advanced prediction of contact force; This provides the timing data of the pantograph-catenary contact force within the current moment and the historical sliding window. For gated cyclic unit nonlinear mapping function; For network weight parameters; The optimal combination of hyperparameters obtained by IGWO optimization.

[0054] Furthermore, step S4 specifically includes: constructing a straddle-type monorail pantograph active control system and performing multi-objective rolling optimization of the objective function.

[0055] Let the prediction time domain be Control time domain is The control objective is to stabilize the pantograph-catenary contact force at the optimal reference value. (Based on straddle-type monorail standard). Rolling optimization objective function Defined as:

[0056]

[0057] In the formula, To predict the length of the time domain; To control the length of the time domain; For ideal reference pantograph-catenary contact force; To control the increment of force; The segmented tracking error penalty weighting coefficient is set based on the contact force hazard threshold. This is the control increment penalty weighting coefficient used to suppress high-frequency, high-amplitude output.

[0058] Furthermore, step S5 specifically includes: under the conditions of hard physical constraints such as the maximum output pull / push force of the actuator, the limit of the actuation response rate, and the range of safe contact force without offline operation, using the sequential quadratic programming method (SQP) to apply the objective function. Perform a fast solution.

[0059] The optimal control increment sequence is obtained by solving. Then, extract the first element. The actual active control force issued is:

[0060]

[0061] In the next control cycle The pantograph-catenary state response after the application of control force is collected, and the error between the actual contact force and the predicted value is calculated. The feedback is fed back into the IGWO-GRU prediction model to eliminate accumulated errors, completing the feedback loop of "perception-prediction-decision-execution".

[0062] The beneficial effects of this invention are: it reduces the fluctuation of pantograph-catenary contact force, improves the current collection quality of the train, and extends the service life of the pantograph contact plate.

[0063] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0064] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0065] Figure 1Schematic diagram of sensor installation position for straddle-type monorail pantograph-catenary contact force acquisition;

[0066] Figure 2 This is a two-dimensional pantograph model;

[0067] Figure 3 This is a schematic diagram of the XGBoost structure;

[0068] Figure 4 This is a schematic diagram of a GRU network;

[0069] Figure 5 System flowchart for active control of straddle-type monorail pantograph;

[0070] Figure 6 A comparison chart of control results at different speeds;

[0071] Figure 7 This is a flowchart of the active control method for straddle-type monorail pantograph based on pantograph-catenary contact force prediction according to the present invention. Detailed Implementation

[0072] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0073] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0074] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0075] like Figure 1-5 , Figure 7 As shown, an active control method for straddle-type monorail pantographs for predicting pantograph-catenary contact force is presented.

[0076] 1) Obtain the pantograph-catenary system status characteristic data and actual pantograph-catenary contact force data during the operation of straddle-type monorail trains;

[0077] 2) Construct an intelligent model of a straddle-type monorail pantograph-catenary system driven by both "mechanism and data" and use the PSO-XGBoost algorithm to accurately compensate for the nonlinear residuals of the pantograph-catenary binary quality model.

[0078] 3) Extract the timing features of the pantograph-catenary contact force and input them into the IGWO-GRU neural network to construct a straddle-type single-track pantograph-catenary contact force prediction model, so as to achieve accurate prediction of the pantograph-catenary contact force in the next control cycle.

[0079] 4) Construct a nonlinear model predictive control framework for straddle-type monorail pantographs. Use the IGWO-GRU straddle-type monorail pantograph-catenary contact force prediction model as a nonlinear predictive operator, and combine the reference pantograph-catenary contact force with the actuator control increment to construct a rolling optimization objective function.

[0080] 5) Under the physical constraints of the straddle-type monorail pantograph-catenary system, the rolling optimization objective function is solved online using the Sequential Quadratic Programming (SQP) method to obtain the optimal control increment sequence; the first control increment of the optimal control increment sequence is extracted and applied to the pantograph actuator, and a closed-loop correction mechanism is constructed.

[0081] First, it is necessary to obtain the contact force data of the straddle-type monorail pantograph-catenary system. The attached diagram illustrates the specific process of this active control system.

[0082] The following example illustrates this embodiment:

[0083] Step 1: Obtain the pantograph-catenary system status characteristics data and actual pantograph-catenary contact force data during the operation of the straddle-type monorail train.

[0084] Data on pantograph acceleration, displacement, temperature, and grating sensor wavelength were acquired during actual vehicle testing on Chongqing Rail Transit Line 3. An OSC7100 Bragg grating accelerometer from Beijing Tongwei Technology, with a sensitivity of 0.17g, was selected. The principle is to calculate acceleration by detecting changes in the center wavelength. Parameters were set as follows: wavelength range 1529–1564 nm, wavelength resolution 1 pm, and scanning frequency 1 kHz. The optical signal was then processed by a demodulator and converted into an acceleration signal. Finally, the pantograph-catenary contact force was calculated based on the measured acceleration data and the technical parameters of the monorail pantograph.

[0085] The second step is to construct an intelligent model of the straddle-type monorail pantograph-catenary system based on the dual-drive of "mechanism-data" and use the PSO-XGBoost algorithm to accurately compensate for the nonlinear residuals of the pantograph-catenary binary quality model.

[0086] The pantograph and overhead contact line are simplified into a single unit coupled together by elastic elements.

[0087] Treating the rigid contact wire as a time-varying stiffness system, the formula is as follows:

[0088]

[0089] Where: The average stiffness coefficient within the span, i.e. ;in , These are the maximum and minimum stiffness coefficients of the overhead contact system within one span. The stiffness variation coefficient of the overhead contact line is... ; The locomotive's operating speed; The contact wire span; This refers to the runtime.

[0090] The dynamic equation for a straddle-type monorail is:

[0091]

[0092] In the formula: The equivalent stiffness of the overhead contact line; , , The equivalent mass, equivalent damping, and vertical displacement of the bow head; , , The equivalent mass, equivalent damping, and vertical displacement of the frame; , The equivalent stiffness of the bow head support rubber and the lifting spring; This is the static lifting force of the pantograph; This is a vehicle body disturbance signal; This is an excitation signal for a rigid overhead contact line.

[0093] Transform the dynamic equations described in step S21 into state equations;

[0094] Let the state vector be The state-space equations are as follows:

[0095]

[0096] The coefficient matrix in the formula is as follows:

[0097]

[0098]

[0099]

[0100]

[0101] The formula for calculating the contact force between the pantograph and the catenary is as follows:

[0102]

[0103] residual contact force of pantograph-catenary Defined as:

[0104]

[0105] XGBoost learns the error evolution of the pantograph-catenary dual-mass model, enabling high-precision and high-efficiency dynamic compensation for the lack of modeling in the straddle-mounted monorail pantograph-catenary dual-mass model. (Palm-catenary contact force residual output function) (i.e., contact force residual) ) is represented as:

[0106]

[0107] In the formula: Let be the function space of all possible decision trees; For the first The prediction weight of each tree; This refers to the current motion state of the straddle-type monorail pantograph-catenary system (such as vehicle speed, instantaneous acceleration of the pantograph head, displacement, etc.).

[0108] The Particle Swarm Optimization (PSO) algorithm is used to adaptively optimize hyperparameters of XGBoost, such as learning rate, maximum tree depth, and regularization coefficient.

[0109] set up This is the best position the particle has experienced so far. This represents the globally optimal position for the entire particle swarm. In the... In the next iteration, the particle's velocity and position are updated using the following formulas:

[0110]

[0111]

[0112] Reconstructed dynamic pantograph-catenary contact force for:

[0113]

[0114] Step 3: Extract the timing features of the pantograph-catenary contact force and input them into the IGWO-GRU neural network to construct a straddle-type single-track pantograph-catenary contact force prediction model, achieving accurate and advanced prediction of the pantograph-catenary contact force for the next control cycle. The working principle of GRU is as follows:

[0115] set up The input feature vector at time t is , The hidden layer state at time t is The signal propagation mechanism inside the GRU unit is as follows:

[0116] Reset door : Determine the hidden state of history How much information needs to be forgotten? The closer the reset gate value is to 0, the more historical information is discarded. The calculation formula is:

[0117]

[0118] The straddle-type monorail uses a rigid contact wire, and the greatest external disturbance is the high-frequency mechanical collision caused by the pantograph passing at high speed through "hard points" such as positioners and busbar connectors. When the pantograph suddenly impacts a hard point (current input...), When a sudden change occurs, the system's previous steady-state sliding inertia ( For predicting future contact forces, this becomes a kind of "drag noise." At this point, the neural network, through the calculation of the weight matrix, will reset the gate. The value rapidly approaches 0. This forces the model to discard historical low-frequency smoothing inertia, effectively extracting the current... The transient impact characteristics introduced greatly eliminate the "prediction phase lag" phenomenon at hard points in the MPC prediction operator. Under steady-state conditions, the current input... No mutation occurred, reset gate The value will approach 1. The model preserves historical hidden states to the greatest extent possible. The characteristics in the model ensure that it can output extremely smooth bow-catenary contact forces that conform to Newton's law of inertia in non-abrupt regions.

[0119] Update Gate : Control the current hidden state How much historical information needs to be retained, and how much new information about candidate latent states needs to be received? After processing by the Sigmoid function σ, the output value is mapped to the interval [0,1]. The formula is as follows:

[0120]

[0121] Under stable operating conditions, replace the door. This will imbue the model with historical information. Greater weighting. This "heavy on history, light on mutations" allocation strategy is equivalent to building a virtual "low-pass filter" at the algorithm's underlying layer. It effectively suppresses high-frequency white noise caused by irrelevant interference, ensuring that the contact force curve output by the model conforms to the fundamental constraint of Newtonian inertial continuity. When the pantograph suddenly strikes a hard point, the low-frequency steady state of the pantograph-catenary system is instantly broken. Faced with extreme input characteristics... , update door It will quickly adjust the output of the weight matrix, significantly reducing historical stationary inertia. Instead of receiving new information about candidate latent states containing disturbances, the proportion of [the data] is shifted.

[0122] Candidate hidden state : Reset the historical information after gate filtering and compare it with the current input To splice, through The activation function scales the data to the range [-1, 1], generating candidate features for the current time step:

[0123]

[0124] Candidate hidden state The generation process is essentially the process of constructing a virtual contact force response using a predictive model. It couples historical inertia with the physical field of real-time impact through feature stitching and utilizes... The function introduces nonlinear saturated boundary conditions that conform to the actual mechanical characteristics, ensuring that the output pantograph-catenary contact force of the prediction model is always within a reasonable boundary.

[0125] Current hidden state Implicit state is achieved through the update gate. Achieve historical state and candidate states Nonlinear fusion is performed to complete one iteration of time-series information:

[0126]

[0127] Current hidden state It contains the most complete historical pantograph-catenary contact force characteristics of the pantograph-catenary system at time t. It will be passed to the next time step t+1 as extremely critical prior information and finally mapped to the contact force prediction scalar of the output layer. High-fidelity iteration ensures that the GRU network can accurately fit and sense the changing trend of the bow-catenary contact force.

[0128] In the above formula: , , These are the weight matrices for the corresponding gates; , , These are the corresponding bias vectors; Use the Sigmoid activation function; This represents the element-wise product of matrix elements.

[0129] An improved Grey Wolf Optimization (IGWO) algorithm is used to optimize the hyperparameters of the GRU: a Logistic chaotic mapping is introduced to generate a uniform initial population; a nonlinear dynamic convergence factor is introduced to dynamically adjust the step size of global exploration and local exploitation.

[0130] The logistic mapping strategy initializes the wolf pack, and its iterative formula is:

[0131]

[0132] In the formula: These are the parameters for controlling chaos. Let be the value of the k-th chaotic sequence of the i-th individual in the d-dimensional space.

[0133] Mapping chaotic sequences to the actual hyperparameter search space of IGWO:

[0134]

[0135] In the formula: , Let be the upper and lower bounds of the parameters in the d-th dimension of the search space.

[0136] The chaotic sequences generated by introducing chaotic mapping have ergodicity, aperiodicity, and randomness, which can make the initial population more evenly cover the entire search space.

[0137] The mathematical expression of the nonlinear decreasing convergence factor strategy can be reconstructed as follows:

[0138]

[0139] In the formula: The current iteration number This represents the maximum number of iterations.

[0140] After adopting this nonlinear convergence factor, in the initial stage of iteration, The decay is relatively slow, and the algorithm can maintain a relatively long period. This allows for a thorough global exploration, significantly reducing the probability of getting trapped in local optima; in the later stages of iteration, The rapid decay to 0 prompts the wolf pack to converge quickly to the optimal solution, significantly enhancing its local fine-grained development capabilities.

[0141] The mathematical model for the location update behavior of an individual gray wolf surrounding and tracking its prey is as follows:

[0142]

[0143]

[0144]

[0145] The optimized output equation for pantograph-catenary contact force prediction by IGWO is:

[0146]

[0147] In the formula, In the current Always looking towards the future Advanced prediction of contact force; This provides the timing data of the pantograph-catenary contact force within the current moment and the historical sliding window. For gated cyclic unit nonlinear mapping function; For network weight parameters; The optimal combination of hyperparameters obtained by IGWO optimization.

[0148] Step 4: Construct a nonlinear model predictive control framework for straddle-type monorail pantographs. The IGWO-GRU straddle-type monorail pantograph-catenary contact force prediction model is used as a nonlinear prediction operator. Combined with the reference pantograph-catenary contact force and actuator control increment, a rolling optimization objective function is constructed.

[0149] Let the active control force be Assuming it acts directly on the mass of the pantograph base frame The dynamic equations of the straddle-type monorail pantograph-catenary system are corrected as follows:

[0150]

[0151] Straddle-type monorail pantograph-catenary system is subject to Under the influence of the action, the state-space equations are modified to:

[0152]

[0153] In the formula:

[0154]

[0155] Dynamic pantograph-catenary contact force The PSO-XGBoost straddle-type monorail pantograph-catenary system intelligent model constructed in Chapter 2 is used for reconstruction:

[0156]

[0157] Based on the railway standard TB / T3271—2011 and the experience of straddle-type monorail operation in Chongqing, the optimal fluctuation range of the pantograph-catenary contact force for straddle-type monorails is between 59±10N, while a contact force between 44 and 79N can ensure normal current collection by the pantograph-catenary system. Therefore, an ideal reference pantograph-catenary contact force is adopted. .

[0158] Input the current and historical characteristics of the straddle-type monorail pantograph-catenary system. Future control capabilities to be optimized:

[0159]

[0160] objective function for:

[0161]

[0162] In the formula: For the first The error penalty weighting coefficient for each step; the larger the value, the lower the system's tolerance for fluctuations in pantograph-catenary contact force.

[0163]

[0164] For the first The step-by-step control increment penalty weight coefficient is used to suppress high-frequency, high-amplitude control outputs.

[0165]

[0166] Within the rolling optimization window, apply the following constraints:

[0167]

[0168] Prevent exceeding the actuator output range.

[0169]

[0170] Limit the actuator response amplitude.

[0171]

[0172] Prevent exceeding the maximum contact force.

[0173]

[0174] Step 5: Under the physical constraints of the straddle-type monorail pantograph-catenary system, the rolling optimization objective function is solved online using the Sequential Quadratic Programming (SQP) method to obtain the optimal control increment sequence; the first control increment of the optimal control increment sequence is extracted and applied to the pantograph actuator, and a closed-loop correction mechanism is constructed.

[0175] The optimal control increment sequence is obtained by online rolling solution of the constrained objective function using the Sequential Quadratic Programming (SQP) method.

[0176]

[0177] The "rolling optimization" mechanism of model predictive control extracts the first element of the optimal sequence. The active control force of the pantograph is transmitted to the pantograph actuator:

[0178]

[0179] In the next control cycle The pantograph-catenary state response after the application of control force is collected, and the error between the actual contact force and the predicted value is calculated. The feedback is fed back into the IGWO-GRU prediction model to eliminate accumulated errors, completing the feedback loop of "perception-prediction-decision-execution".

[0180] observe Figure 6 The advantages of this invention are demonstrated through experimental comparison. To further verify the effectiveness and superiority of the proposed method, it is compared with PID control and MPC control, which currently has better overall performance, at different speeds. Simulation experiments are conducted using Simulink based on the state-space equations.

[0181] The formula for calculating the PID control force is as follows:

[0182] In the formula: , , These are the proportional gain, integral gain, and differential gain coefficients, respectively. These three gain coefficients are determined using a trial-and-error method based on engineering experience.

[0183] The formula for calculating the MPC control force is as follows:

[0184] In the formula This refers to the initial control force (i.e., the basic control state) at the previous moment. This represents the optimal control force increment at the current moment.

[0185] In summary, the PID controller and MPC controller have been successfully built.

[0186] Current collection quality evaluation. To analyze current collection quality, the maximum value is selected. Minimum value ,average value The standard deviation (SD) is used as a metric.

[0187] The selected parameters are shown in Table 1, and the comparison results are shown in Table 2.

[0188] Table 1. Pantograph-Catenary Parameters

[0189]

[0190] Table 2 Comparison of pantograph-catenary contact force control results at different speeds

[0191]

[0192] To improve the current collection quality of vehicles and extend the service life of the pantograph sliding plate, this invention proposes an active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction.

[0193] First, contact force data of the pantograph-catenary system is acquired. Then, an intelligent model of the straddle-type monorail pantograph-catenary system based on a dual-driven "mechanism-data" approach is established. Next, a pantograph-catenary contact force prediction model based on the IGWO-GRU is constructed to accurately predict the contact force at the next time step, providing a more reliable predicted contact force input for active pantograph control. Finally, an active control model for the straddle-type monorail pantograph based on pantograph-catenary contact force prediction is built, overcoming the problems of existing control methods relying on expert experience and having low reliability. This model fully leverages the advantages of the straddle-type monorail pantograph-catenary system, such as data-driven operation, adaptive optimization of model parameters, and high anti-interference capability of the control model. Its advantages are verified by comparison with other methods.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction, characterized in that: Includes the following steps: S1: Obtain the pantograph-catenary system status characteristic data and actual pantograph-catenary contact force data during the operation of straddle-type monorail trains; S2: Construct an intelligent model of a straddle-type monorail pantograph-catenary system based on the dual-drive of "mechanism-data" and use the PSO-XGBoost algorithm to accurately compensate for the nonlinear residuals of the pantograph-catenary binary quality model; S3: Extract the timing features of the pantograph-catenary contact force and input them into the IGWO-GRU neural network to construct a straddle-type single-track pantograph-catenary contact force prediction model, so as to achieve accurate prediction of the pantograph-catenary contact force in the next control cycle. S4: Construct a nonlinear model predictive control framework for straddle-type monorail pantographs, using the IGWO-GRU straddle-type monorail pantograph-catenary contact force prediction model as a nonlinear predictive operator, and combining the reference pantograph-catenary contact force with the actuator control increment to construct a rolling optimization objective function; S5: Under the physical constraints of the straddle-type monorail pantograph-catenary system, the rolling optimization objective function is solved online using the Sequential Quadratic Programming (SQP) method to obtain the optimal control increment sequence; the first control increment of the optimal control increment sequence is extracted and applied to the pantograph actuator, and a closed-loop correction mechanism is constructed.

2. The active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction according to claim 1, characterized in that, In S2, the specific process of constructing an intelligent model for a straddle-type monorail pantograph-catenary system based on a dual-driven "mechanism-data" approach includes: S21: Establish a two-dimensional mass dynamics model of the straddle-type monorail pantograph-catenary coupling system and calculate the reference pantograph-catenary contact force constrained by physical and mechanical laws; S22: Construct an extreme gradient boosting tree (XGBoost) model, taking the state characteristics of the straddle-type monorail pantograph-catenary system as input, and the residual between the actual pantograph-catenary contact force and the reference pantograph-catenary contact force as the fitting target; S23: The Particle Swarm Optimization (PSO) algorithm is used to perform global adaptive optimization of the core hyperparameters of the XGBoost model, and outputs nonlinear residual compensation terms; The dynamic pantograph-catenary contact force is reconstructed by superimposing the reference pantograph-catenary contact force with the nonlinear residual compensation term.

3. The active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction according to claim 1, characterized in that, In S3, the construction process of the IGWO-GRU straddle-type monorail pantograph-catenary contact force prediction model includes: S31: A Logistic chaotic mapping strategy is used to generate a uniformly distributed initial gray wolf population; a nonlinear dynamic convergence factor based on the cosine function is introduced to update the position of individual gray wolves, demonstrating the global exploration and local development capabilities of the dynamic equilibrium algorithm. S32: Using the root mean square error of the predictions of the GRU neural network on the validation set as the fitness function, the improved gray wolf optimization algorithm (IGWO) is used to globally optimize the number of hidden layer nodes, the initial learning rate, and the regularization coefficient of the GRU neural network. S33: The globally optimal hyperparameter combination is assigned to the GRU network to form a nonlinear prediction model for the contact force of the straddle-type monorail pantograph-catenary system.

4. The active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction according to claim 3, characterized in that, The IGWO-GRU prediction operator uses a sliding time window to extract time-series samples of the pantograph-catenary contact force and performs multi-step prediction within the rolling time domain. Its pantograph-catenary contact force prediction output equation is as follows: In the formula, In the current Always looking towards the future Advanced prediction of contact force; This provides the timing data of the pantograph-catenary contact force within the current moment and the historical sliding window. For gated cyclic unit nonlinear mapping function; For network weight parameters; The optimal combination of network hyperparameters obtained by IGWO optimization.

5. The active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction according to claim 1, characterized in that, In S4, the rolling optimization objective function is defined as a multi-objective penalty function for the control increment to be optimized and the tracking error: In the formula, To predict the length of the time domain; To control the length of the time domain; For ideal reference pantograph-catenary contact force; To control the increment of force; The segmented tracking error penalty weighting coefficient is set based on the contact force hazard threshold. This is the control increment penalty weighting coefficient used to suppress high-frequency, high-amplitude output.

6. The active control method for straddle-type monorail pantographs based on pantograph-catenary contact force prediction according to claim 1, characterized in that, In S5, the specific execution logic of the closed-loop correction mechanism is as follows: pantograph actuator in Actual active control at all times It is obtained by adding the control force from the previous moment to the extracted first optimal control increment; exist At any given moment, the actual dynamic response of the pantograph-catenary system after the application of control force is collected, and the feedforward prediction error between the actual contact force and the predicted value at the previous moment is calculated. The latest state feature matrix, which includes the straddle-type monorail pantograph-catenary system and the feedforward error, is re-input into the IGWO-GRU prediction model to achieve online rolling correction of the prediction cumulative error.

Images