A method of adaptive speed control of a train

Abstract

The application discloses a kind of adaptive speed control method of train, belong to train speed control technical field.The application is by the deviation sequence of actual speed and target speed in historical time period, the time accumulation of positive and negative part is carried out, and slow-time environmental disturbance and short-time environmental disturbance are obtained respectively;According to two kinds of disturbance, the corresponding environmental disturbance model is established, slow-time compensation and short-time compensation are obtained, and they are introduced into outer layer PID model to form outer layer PID compensation model;Further combined with predictive feedforward control quantity, inner layer PID model is constructed, and the rapid response to speed change is realized;Again according to historical target speed and inner layer PID output, adaptive adjustment is carried out to current control output, and the final speed control quantity is obtained.The application is combined by multilayer nested PID and adaptive compensation mechanism, can effectively inhibit speed deviation accumulation and response lag, improve the speed control precision and stability of train under complex working condition.

Description

Technical Field

[0001] This invention relates to the field of train speed control technology, and more specifically to an adaptive speed control method for trains. Background Technology

[0002] Currently, the field of train speed control widely adopts a control architecture based on PID control. This architecture uses the synergistic action of proportional, integral, and derivative components to regulate the traction and braking systems, achieving tracking of the target speed curve. To improve control performance, the industry has developed improved PID schemes that combine fuzzy control and iterative learning strategies, attempting to enhance the system's adaptability to environmental changes through parameter optimization. Simultaneously, the application of programmable logic controllers (PLCs) further improves the efficiency of control command execution, and, in conjunction with sensor networks, enables real-time acquisition and processing of speed signals, providing data support for control decisions.

[0003] However, existing speed control technologies still have significant limitations in dealing with environmental disturbances: traditional PID control and its improvement schemes mostly rely on fixed control parameters or single-dimensional disturbance compensation mechanisms, which cannot effectively adapt to the two types of environmental disturbances commonly encountered in train operation—slow time-varying disturbances such as gradual changes in track gradient and long-term load changes, and short-term transient disturbances such as sudden airflow and track joint impacts. This lack of identification of disturbance characteristics leads to a lack of targeted compensation strategies. When the two types of disturbances occur simultaneously, the control system is prone to accumulated speed deviations or response lags, making it difficult to maintain a precise match between the actual speed and the target speed.

[0004] More importantly, existing technologies that rely solely on single-layer PID control based on the difference between the target speed and the actual speed suffer from several problems: First, control accuracy is easily affected by system delays, and the single-layer structure cannot handle the needs of "target tracking" and "interference suppression" in a layered manner, leading to overshoot or oscillation during speed fine-tuning. Second, parameter adaptability is poor; the proportional, integral, and derivative parameters of a single-layer PID controller need to consider different operating conditions, making it difficult to find the optimal balance between "fast response" and "steady-state stability." For example, increasing the proportional coefficient to cope with sudden disturbances can easily lead to aggravated speed fluctuations. Third, disturbance rejection robustness is insufficient; when operating conditions such as line load and external resistance change abruptly, single-layer control cannot quickly isolate the disturbance source and compensate accordingly, relying only on error feedback for passive adjustment, further amplifying speed deviations and even affecting train operation safety and comfort. Summary of the Invention

[0005] In view of the above-mentioned shortcomings in the prior art, the adaptive speed control method for trains provided by the present invention solves the problems of the existing control system being prone to speed deviation accumulation or response lag, difficulty in maintaining accurate matching between actual speed and target speed, and the shortcomings of single-layer PID control.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: an adaptive speed control method for trains, comprising the following steps:

[0007] S1. Accumulate the positive and negative parts of the speed deviation sequence between the actual and target speeds in each historical time period to obtain the slow-time environmental interference and short-time environmental interference based on the cumulative difference between the positive and negative values.

[0008] S2. Based on the slow-time environmental interference amount, set up a slow-time environmental interference model to obtain the slow-time compensation amount. Based on the short-time environmental interference amount, set up a short-time environmental interference model to obtain the short-time compensation amount.

[0009] S3. Add the slow-time compensation amount and the short-time compensation amount to the outer PID model to obtain the outer PID compensation model;

[0010] S4. Predict the feedforward control quantity and construct the inner PID model based on the output of the outer PID compensation model and the feedforward control quantity.

[0011] S5. Obtain the adaptation coefficient based on the historical target speed and the output of the inner PID model;

[0012] S6. Adjust the output of the current inner-layer PID model according to the adaptation coefficient to obtain the speed control quantity.

[0013] Furthermore, S1 includes the following sub-steps:

[0014] S11. Within each historical time period, calculate the difference between the historical actual speed and the historical target speed to obtain the speed deviation sequence;

[0015] S12. Integrate the portion of the velocity deviation sequence that is greater than zero over the sampling interval to obtain the cumulative positive deviation; integrate the portion that is less than zero over the sampling interval to obtain the cumulative negative deviation.

[0016] S13. Subtract the cumulative positive deviation from the cumulative negative deviation to obtain the cumulative difference for each historical time period;

[0017] S14. Take the average of the cumulative differences over multiple historical time periods and use the average as the slow-time environmental disturbance quantity.

[0018] S15. Use the cumulative difference of the latest historical time period as the short-term environmental disturbance quantity.

[0019] Furthermore, the process of setting the slow-time environmental disturbance model in S2 includes: multiplying the adjustable parameter of the slow-time environmental disturbance quantity with the slow-time environmental disturbance quantity, and mapping the multiplication result into a slow-time compensation quantity through a mapping function;

[0020] The process of setting up a short-term environmental disturbance model includes: multiplying the adjustable parameter of the short-term environmental disturbance quantity with the short-term environmental disturbance quantity, and mapping the multiplication result into a short-term compensation quantity through a mapping function.

[0021] Furthermore, the input to the outer PID compensation model in S3 is the outer loop error, which is equal to the difference between the target speed and the actual speed.

[0022] The outer-layer PID compensation model includes: the proportional term of the outer loop error, the integral term of the outer loop error, the differential term of the outer loop error, the slow-time compensation term, and the short-time compensation term.

[0023] Furthermore, S4 includes the following sub-steps:

[0024] S41. Based on the dynamic model, predict the feedforward control quantity:

[0025] S42. Add the feedforward control quantity to the output of the outer PID compensation model and subtract the actual speed to obtain the feedforward fusion inner loop error.

[0026] S43. Set the proportional term, piecewise integral term, and band-limited derivative term for the feedforward fusion inner loop error to obtain the inner layer PID model.

[0027] Furthermore, the expression for the inner PID model in S43 is: ,

[0028] Among them, u 2,t Let e ​​be the output of the inner PID model at time t. 2,t Let K be the feedforward fusion inner loop error at time t. p2 K is the proportionality coefficient of the inner loop error. i2 K represents the integral coefficient of the inner loop error. d2 I is the differential coefficient of the inner loop error. seg,t For the piecewise integral at time t, D(e 2,t Let be the band-limited differential at time t.

[0029] Furthermore, piecewise integral I seg,t In the feedforward fusion inner loop error e 2,t If the direction of the slow-time compensation is consistent with that of the slow-time compensation, and the absolute value of the slow-time compensation is less than the allowable threshold for the slow-time compensation, then normal integration occurs. Otherwise, an attenuation factor is set, and the attenuation factor is compared with the piecewise integral I of the previous time step. seg,t-1 Multiply, and use the result of the multiplication as the current piecewise integral I. seg,t ;

[0030] The band-limited differential D(e) at time t 2,t The calculation process includes: at the current time t, obtaining the feedforward fusion inner loop error e. 2,tand the feedforward fusion inner loop error e from the previous time step 2,t-1 The rate of change of error is obtained, and the band-limited differential D(e) of the previous time step is read. 2,t-1 Based on the filtering time constant and sampling period, calculate the filtering coefficients, and then apply the band-limited differential D(e^(-1 / 2)) to the sampled area. 2,t-1 The band-limited differential D(e^t) at time t is obtained by proportionally integrating the error rate of change with the error rate. 2,t ).

[0031] Furthermore, S5 includes the following sub-steps:

[0032] S51. Calculate the ratio of target velocities at adjacent time points to obtain the target velocity change ratio;

[0033] S52. Calculate the output ratio of the inner-layer PID model at adjacent time points to obtain the control quantity change ratio;

[0034] S53. Calculate the adaptation coefficient at time t based on the ratio of the change in target velocity to the ratio of the change in control quantity.

[0035] Furthermore, the formula for calculating the fitness coefficient at time t in S53 is as follows: ,

[0036] Where, ζ t Let be the fitness coefficient at time t, exp be the exponential function, and r be the coefficient of fitness. v,t r is the ratio of the target velocity change at time t. u,t Let be the change ratio of the control quantity at time t, ln be the logarithmic function, k be the trend sensitivity coefficient, and || be the absolute value operation.

[0037] Furthermore, the expression for the speed control quantity obtained in S6 is: ,

[0038] Among them, u 2,t,z Let ζ be the velocity control variable at time t. t Let u be the fitness coefficient at time t. 2,t u is the output of the inner PID model at time t. 2,t-1 This represents the output of the inner PID model at time t-1, where t is the time number.

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

[0040] 1. This invention accumulates the positive and negative parts of the historical speed deviation sequence over time, which can accurately distinguish and quantify slow time-varying interference and short-term transient interference. Combined with step S2, it constructs models for the two types of interference and generates corresponding compensation amounts, which can specifically offset the effects of interference with different characteristics such as gradual changes in line gradient and sudden airflow, avoiding the insufficient adaptability caused by the "one-size-fits-all" approach of traditional single compensation mechanisms, and ensuring accurate matching between actual speed and target speed.

[0041] 2. This invention adopts a dual-layer PID model architecture to overcome the limitations of single-layer control, significantly improving control accuracy and parameter adaptability. Step S3 integrates two types of compensation quantities into the outer PID to form a compensation model, which is responsible for prioritizing the "interference suppression" requirement; Step S4 combines the inner PID model constructed with the feedforward control quantity to focus on achieving "target speed tracking". The layered design allows the two models to perform their respective functions, without having to consider multiple requirements as in a single-layer PID. It can quickly cancel interference through the outer compensation and precisely fine-tune the speed through the inner layer, effectively avoiding overshoot and oscillation problems, while avoiding the contradiction between "fast response" and "steady-state stability" during parameter tuning.

[0042] 3. This invention enhances system robustness through a dynamic adjustment mechanism of the adaptation coefficient, ensuring operational safety and comfort under complex conditions. Step S5 calculates the adaptation coefficient based on historical target speed and inner PID output, which can capture the changing trend of train operating conditions in real time. Step S6 dynamically adjusts the current control quantity using this coefficient, enabling the system to quickly adapt to changes in operating conditions under complex scenarios such as sudden load changes and resistance fluctuations. This avoids the amplification of deviations caused by traditional passive error feedback, not only improving the system's anti-disturbance capability but also reducing the impact of speed fluctuations on train operation safety and comfort, providing a reliable guarantee for stable train operation. Attached Figure Description

[0043] Figure 1 A flowchart of an adaptive speed control method for trains;

[0044] Figure 2 A graph showing the comparison between the train's target speed and actual speed;

[0045] Figure 3 This is a graph showing the absolute value of the tracking error. Detailed Implementation

[0046] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, an adaptive speed control method for trains includes the following steps:

[0048] S1. Accumulate the positive and negative parts of the speed deviation sequence between the actual and target speeds in each historical time period to obtain the slow-time environmental interference and short-time environmental interference based on the cumulative difference between the positive and negative values.

[0049] S2. Based on the slow-time environmental interference amount, set up a slow-time environmental interference model to obtain the slow-time compensation amount. Based on the short-time environmental interference amount, set up a short-time environmental interference model to obtain the short-time compensation amount.

[0050] S3. Add the slow-time compensation amount and the short-time compensation amount to the outer PID model to obtain the outer PID compensation model;

[0051] S4. Predict the feedforward control quantity and construct the inner PID model based on the output of the outer PID compensation model and the feedforward control quantity.

[0052] S5. Obtain the adaptation coefficient based on the historical target speed and the output of the inner PID model;

[0053] S6. Adjust the output of the current inner-layer PID model according to the adaptation coefficient to obtain the speed control quantity.

[0054] In this embodiment, S1 includes the following sub-steps:

[0055] S11. Within each historical time period, calculate the difference between the historical actual speed and the historical target speed to obtain the speed deviation sequence;

[0056] S12. Integrate the portion of the velocity deviation sequence that is greater than zero over the sampling interval to obtain the cumulative positive deviation; integrate the portion that is less than zero over the sampling interval to obtain the cumulative negative deviation (i.e., accumulate the product of the deviation value and the sampling interval at the sampling time).

[0057] S13. Subtract the cumulative positive deviation from the cumulative negative deviation to obtain the cumulative difference for each historical time period;

[0058] S14. Take the average of the cumulative differences over multiple historical time periods and use the average as the slow-time environmental disturbance quantity.

[0059] S15. Use the cumulative difference of the latest historical time period as the short-term environmental disturbance quantity.

[0060] In this embodiment, the length of a historical time period is 5 seconds. Ten historical time periods are selected, and the average of the cumulative differences of the ten historical time periods is taken. A speed deviation sequence contains 50 differences.

[0061] Subtracting the historical actual speed from the historical target speed at the same time point yields multiple speed differences, which are then combined to form a speed deviation sequence.

[0062] This invention constructs a speed deviation sequence based on the difference between target and actual speeds over a historical time period, and integrates the positive and negative parts of the deviation over time to distinguish the cumulative trend of actual speed being higher or lower than the target speed during train operation. By averaging the cumulative differences over multiple historical time periods, a slow-time environmental disturbance quantity reflecting the long-term operating state is obtained, which can be used to characterize slowly changing factors such as track gradient changes and load fluctuations. Simultaneously, the cumulative difference of the most recent time period is used as a short-time environmental disturbance quantity to reflect the short-term effects of instantaneous disturbances such as wind resistance fluctuations and adhesion changes. This method achieves the separation and quantification of environmental disturbances at different time scales, enabling the control system to maintain long-term speed stability while quickly responding to short-time disturbances, thereby improving the accuracy, smoothness, and anti-interference capability of train speed control.

[0063] In this embodiment, the process of setting the slow-time environmental interference model in S2 includes: multiplying the adjustable parameter of the slow-time environmental interference amount with the slow-time environmental interference amount, and mapping the multiplication result into a slow-time compensation amount through a mapping function;

[0064] The process of setting up a short-term environmental disturbance model includes: multiplying the adjustable parameter of the short-term environmental disturbance quantity with the short-term environmental disturbance quantity, and mapping the multiplication result into a short-term compensation quantity through a mapping function.

[0065] The expression for the slow-time environmental disturbance model is: ,

[0066] Among them, C slow α is the slow-time compensation amount, tanh is the mapping function, and α is the slow-time compensation amount. slow S is an adjustable parameter for the amount of environmental disturbance in slow time. slow This represents the amount of environmental disturbance during slow-motion operation.

[0067] The expression for the short-time environmental disturbance model is: ,

[0068] Among them, C short For short-term compensation, α short S is an adjustable parameter for short-term environmental disturbances. short This refers to short-term environmental disturbances.

[0069] In this embodiment, α slow Set within the range of 0.1 to 0.5, α short Set within the range of 0.5 to 2.0. The specific value can be adjusted according to needs during use.

[0070] In this embodiment, the input of the outer PID compensation model in S3 is the outer loop error, which is equal to the difference between the target speed and the actual speed.

[0071] The outer-layer PID compensation model includes: the proportional term of the outer loop error, the integral term of the outer loop error, the differential term of the outer loop error, the slow-time compensation term, and the short-time compensation term.

[0072] Among them, e 1,t Let r be the outer loop error at time t. t Let y be the target velocity at time t. t The actual velocity at time t;

[0073] The expression for the outer PID compensation model is: ,

[0074] Among them, u 1,t K represents the output of the outer PID compensation model at time t. p1 K is the proportionality coefficient of the outer loop error. i1 K represents the integral coefficient of the outer loop error. d1 denoted by , where β is the differential coefficient of the outer loop error, Υ is the proportional coefficient of the slow-time compensation, τ is the proportional coefficient of the short-time compensation, and τ is the integral parameter. The outer loop error e 1,t The differential, C slow C is the slow time compensation amount. short This is for short-term compensation.

[0075] This invention introduces separate slow-time compensation terms and short-time compensation terms, which can specifically offset environmental interference with different characteristics such as gradual changes in line gradient and sudden airflow. This avoids the passivity of traditional PID control that relies solely on error feedback, reduces the impact of interference on speed control from the source, and significantly reduces the risk of speed deviation accumulation.

[0076] In this embodiment, S4 includes the following sub-steps:

[0077] S41. Based on the dynamic model, predict the feedforward control quantity:

[0078] S42. Add the feedforward control quantity to the output of the outer PID compensation model and subtract the actual speed to obtain the feedforward fusion inner loop error.

[0079] S43. Set the proportional term, piecewise integral term, and band-limited derivative term for the feedforward fusion inner loop error to obtain the inner layer PID model.

[0080] In this embodiment, the formula for the dynamic model in step S41 is: ,

[0081] Among them, u f,t For the predicted feedforward control quantity, For equivalent quality, To predict the target acceleration, The force is the component caused by the slope, and Frolling is the rolling resistance.

[0082] In this embodiment, the expression for the feedforward fusion inner loop error in S42 is: ,

[0083] Among them, e 2,t Let u be the inner loop error of the feedforward fusion at time t. 1,t For the output of the outer PID compensation model, u f,t To predict the feedforward control input, y t Let be the actual velocity at time t.

[0084] In this embodiment, the expression for the inner PID model in S43 is: ,

[0085] Among them, u 2,t Let e ​​be the output of the inner PID model at time t. 2,t Let K be the feedforward fusion inner loop error at time t. p2 K is the proportionality coefficient of the inner loop error. i2 K represents the integral coefficient of the inner loop error. d2 I is the differential coefficient of the inner loop error. seg,t For the piecewise integral at time t, D(e 2,t Let be the band-limited differential at time t.

[0086] In this embodiment, piecewise integral I seg,t In the feedforward fusion inner loop error e 2,t If the direction of the slow-time compensation is consistent with that of the slow-time compensation, and the absolute value of the slow-time compensation is less than the allowable threshold for the slow-time compensation, then normal integration occurs. Otherwise, an attenuation factor is set, and the attenuation factor is compared with the piecewise integral I of the previous time step. seg,t-1 Multiply, and use the result of the multiplication as the current piecewise integral I. seg,t : ,

[0087] Among them, e 2,t Let C be the inner loop error of the feedforward fusion at time t, sign be the sign function, and C be the inner loop error of the feedforward fusion. int C is the allowable threshold for slow-time compensation. slow is the slow compensation amount, and means "and", or means "other cases", and ε is the decay factor.

[0088] In this embodiment, ε is the attenuation factor set within the range of 0.5 to 0.8, and the allowable threshold C for slow time compensation is... int Set to 0.5 times the maximum slow compensation amount.

[0089] The band-limited differential D(e) at time t 2,tThe calculation process includes: at the current time t, obtaining the feedforward fusion inner loop error e. 2,t and the feedforward fusion inner loop error e from the previous time step 2,t-1 The rate of change of error is obtained, and the band-limited differential D(e) of the previous time step is read. 2,t-1 Based on the filtering time constant and sampling period, calculate the filtering coefficients, and then apply the band-limited differential D(e^(-1 / 2)) to the sampled area. 2,t-1 The band-limited differential D(e^t) at time t is obtained by proportionally integrating the error rate of change with the error rate. 2,t ): ,

[0090] Wherein, D(e) 2,t-1 Let be the band-limited differential at time t. T is the filtering time constant. s For the sampling period, e 2,t-1 Let be the feedforward fusion inner loop error at time t-1.

[0091] The filter time constant is set in the range of 0.01 to 0.1 seconds.

[0092] This invention predicts the feedforward control quantity based on a dynamic model and integrates it with the outer PID output to form an inner loop error. This can compensate for known disturbances such as slope force and rolling resistance in advance, enabling the system to apply control at the "source" of speed change, significantly improving dynamic response speed and reducing the lag deviation between actual speed and target speed.

[0093] This invention achieves dynamic coupling between feedback correction and feedforward prediction by adding the output of the outer PID compensation model to the predicted feedforward quantity and subtracting the actual speed to form the feedforward fusion inner loop error. This effectively suppresses speed deviations caused by gradient changes or sudden changes in resistance. Based on this, the inner PID model employs a composite structure of piecewise integral and band-limited derivative: the piecewise integral adaptively adjusts its intensity based on the sign relationship between the error direction and the slow-time compensation quantity, automatically triggering integral decay when the system error direction is inconsistent with the environmental disturbance direction and the disturbance amplitude is too large, thus avoiding integral saturation and oscillation; the band-limited derivative term is fused by the ratio of the filtering time constant to the sampling period, suppressing high-frequency noise interference while maintaining a sensitive response to error changes. This inner control structure significantly improves the system's stability and robustness while ensuring dynamic response speed, enabling train speed control to achieve fast, stable, and accurate tracking even under complex gradient, load, and environmental changes.

[0094] Slow-time compensation corresponds to continuous and stable slow-time-varying disturbances such as gradual changes in track gradient and long-term load changes, while short-time compensation corresponds to immediate and brief short-term transient disturbances such as sudden airflow and track joint impacts. By directly integrating both into the outer PID, targeted compensation can be actively applied in the early stages of disturbances affecting the system, reducing the impact of disturbances on speed control from the source. This avoids the passivity of traditional PID relying solely on error feedback and significantly reduces the risk of speed deviation accumulation.

[0095] The feedforward control quantity is predicted based on the dynamic model and can compensate for known systematic disturbances such as slope force and rolling resistance in advance. It is combined with the outer PID output and the actual speed to form an inner loop error. This allows the inner PID to offset the effects of these predictable disturbances in advance when tracking the target speed, thereby improving the dynamic accuracy and response speed of speed tracking. This achieves the hierarchical control effect of "active prediction + precise tracking", enabling the entire system to maintain accurate matching with the target speed even under complex working conditions.

[0096] In this embodiment, S5 includes the following sub-steps:

[0097] S51. Calculate the ratio of target velocities at adjacent time points to obtain the target velocity change ratio;

[0098] S52. Calculate the output ratio of the inner-layer PID model at adjacent time points to obtain the control quantity change ratio;

[0099] S53. Calculate the adaptation coefficient at time t based on the ratio of the change in target velocity to the ratio of the change in control quantity.

[0100] The target velocity change ratio at time t is the ratio of the target velocity at time t to the target velocity at time t-1; the control quantity change ratio at time t is the ratio of the output of the inner PID model at time t to the output of the inner PID model at time t-1.

[0101] In this embodiment, the formula for calculating the fitness coefficient at time t in S53 is as follows: ,

[0102] Where, ζ t Let be the fitness coefficient at time t, exp be the exponential function, and r be the coefficient of fitness. v,t r is the ratio of the target velocity change at time t. u,t Let be the change ratio of the control quantity at time t, ln be the logarithmic function, k be the trend sensitivity coefficient, and || be the absolute value operation.

[0103] In this embodiment, the trend sensitivity coefficient k is set to 1.5.

[0104] The introduction of the adaptation coefficient gives the system the ability to "adaptively adjust," no longer relying on fixed control parameters to cope with all operating conditions. Whether it's rapid changes in the target speed or fluctuations in the control output, the adaptation coefficient can be dynamically adjusted to balance these changes, significantly enhancing the system's robustness against disturbances in complex and variable environments and ensuring the stability of train operation. Calculating the adaptation coefficient based on the ratio of the target speed to the change in the control output quantifies the matching degree between "speed demand" and "control output." Adjusting the current inner-layer PID output using this coefficient allows the control output to more accurately match changes in the target speed, further reducing the deviation between the actual speed and the target speed and improving the accuracy of speed control.

[0105] In this embodiment, the expression for the speed control quantity obtained in S6 is: ,

[0106] Among them, u 2,t,z Let ζ be the velocity control variable at time t. t Let u be the fitness coefficient at time t. 2,t u is the output of the inner PID model at time t. 2,t-1 This represents the output of the inner PID model at time t-1, where t is the time number.

[0107] When the adaptation coefficient is ≥0.8, the current inner-layer PID output is directly used, ensuring that the system can quickly respond to changes in target speed under stable operating conditions and good adaptability, maintaining accurate speed tracking and avoiding response lag caused by excessive smoothing, thus balancing the requirements of "stability" and "response speed". When the adaptation coefficient is <0.8, a weighted fusion method is used, which can smoothly transition between the current inner-layer PID output and historical outputs, avoiding sudden and large changes in control quantities, effectively reducing the impact of trains caused by sudden changes in control quantities, improving running comfort, and also reducing the risk of wear and tear on the traction and braking systems caused by instantaneous high loads.

[0108] In this embodiment, K p1 K i1 K d1 ,β,Υ,K p2 K i2 K d2 The eight parameters are denoted as chromosome vectors. The specific values ​​of these eight parameters are obtained through calibration using a genetic algorithm. The initial settings for the genetic algorithm are: population size of 60 and generation limit of 150. In the genetic algorithm, the distance between the actual speed and the target speed is used as the cost function.

[0109] In this embodiment, a simulation experiment was set up with a total simulation time of 120s and a sampling period of 0.1s (1200 sampling points in total). The train's running status was as follows: 0–20s: uniform acceleration to 80km / h (22.2m / s); 20–60s: uniform speed operation; 60–80s: uphill (10‰) section, target speed slightly decreasing to 70km / h (19.4m / s); 80–100s: downhill (-8‰) section, target speed slightly increasing to 85km / h (23.6m / s); 100–120s: flat section, returning to 80km / h. Gradient interference was as follows: 0–60s: 0‰ (flat ground); 60–80s: +10‰ uphill; 80–100s: -8‰ downhill; 100–120s: 0‰. Figure 2 As shown, the red line represents the train's actual speed response when using a traditional PID controller. It can be seen that the speed drops significantly in the uphill section (60-80s) and suddenly decreases in the disturbance section (around 95s), exhibiting overall lag and fluctuation. The blue solid line represents the train's actual operating speed after adopting your proposed method of "dual-loop PID + environmental disturbance compensation + predictive feedforward + adaptation coefficient." Compared to the red line, it shows: more stable speed during uphill and disturbance periods; closer tracking of the target speed; and a smoother response with less fluctuation. Compared to traditional PID, this invention can more accurately track the target speed, reduce lag error, and maintain speed stability in complex gradient and disturbance environments.

[0110] like Figure 3 As shown, the red curve represents the absolute value of speed error under traditional PID control, while the blue curve represents the absolute value of speed error under the adaptive speed control method of this invention. The red line (traditional PID) shows obvious error peaks during uphill and disturbance periods (around 60s, 95s, and 80s), indicating that traditional PID control is sensitive to external disturbances and has a delayed response. The blue line (the method of this invention) has a smaller overall error and more stable fluctuations, indicating that the system can quickly correct speed deviations through external disturbance compensation, feedforward prediction, and adaptation coefficient adjustment.

[0111] This invention accumulates the positive and negative parts of the historical speed deviation sequence over time, which can accurately distinguish and quantify slow time-varying interference and short-term transient interference. Combined with step S2, it constructs models for the two types of interference and generates corresponding compensation amounts, which can specifically offset the effects of interference with different characteristics such as gradual changes in line gradient and sudden airflow. This avoids the insufficient adaptability caused by the "one-size-fits-all" approach of traditional single compensation mechanisms and ensures accurate matching between actual speed and target speed.

[0112] This invention employs a dual-layer PID model architecture to overcome the limitations of single-layer control, significantly improving control accuracy and parameter adaptability. Step S3 integrates two types of compensation quantities into the outer PID to form a compensation model, responsible for prioritizing the "interference suppression" requirement; Step S4 combines the inner PID model constructed with feedforward control quantities, focusing on achieving "target speed tracking." The layered design allows each layer to perform its own function, eliminating the need to consider multiple requirements as with a single-layer PID. It can quickly offset interference through outer compensation and precisely fine-tune speed through inner layer, effectively avoiding overshoot and oscillation problems, while also avoiding the contradiction between "fast response" and "steady-state stability" during parameter tuning.

[0113] This invention enhances system robustness through a dynamic adjustment mechanism of the adaptation coefficient, ensuring operational safety and comfort under complex conditions. Step S5 calculates the adaptation coefficient based on historical target speed and inner-layer PID output, enabling real-time capture of changes in train operating conditions. Step S6 dynamically adjusts the current control quantity using this coefficient, allowing the system to quickly adapt to changes in operating conditions under complex scenarios such as sudden load changes and resistance fluctuations. This avoids the amplification of deviations caused by traditional passive error feedback, not only improving the system's anti-disturbance capability but also reducing the impact of speed fluctuations on train operation safety and comfort, providing a reliable guarantee for stable train operation.

[0114] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive speed control method for trains, characterized in that, Includes the following steps: S1. Accumulate the positive and negative parts of the speed deviation sequence between the actual and target speeds in each historical time period to obtain the slow-time environmental interference and short-time environmental interference based on the cumulative difference between the positive and negative values. S1 includes the following steps: S11. Within each historical time period, calculate the difference between the historical actual speed and the historical target speed to obtain the speed deviation sequence; S12. Integrate the portion of the velocity deviation sequence that is greater than zero over the sampling interval to obtain the cumulative positive deviation; integrate the portion that is less than zero over the sampling interval to obtain the cumulative negative deviation. S13. Subtract the cumulative positive deviation from the cumulative negative deviation to obtain the cumulative difference for each historical time period; S14. Take the average of the cumulative differences over multiple historical time periods and use the average as the slow-time environmental disturbance quantity. S15. Use the cumulative difference of the latest historical time period as the short-term environmental disturbance quantity; S2. Based on the slow-time environmental interference amount, set up a slow-time environmental interference model to obtain the slow-time compensation amount. Based on the short-time environmental interference amount, set up a short-time environmental interference model to obtain the short-time compensation amount. The expression for the slow-time environmental disturbance model is: ， Among them, C slow α is the slow-time compensation amount, tanh is the mapping function, and α is the slow-time compensation amount. slow S is an adjustable parameter for the amount of environmental disturbance in slow time. slow This refers to the amount of environmental disturbance during slow-motion operation. The expression for the short-time environmental disturbance model is: ， Among them, C short For short-term compensation, α short S is an adjustable parameter for short-term environmental disturbances. short This refers to short-term environmental disturbances. S3. Add the slow-time compensation amount and the short-time compensation amount to the outer PID model to obtain the outer PID compensation model; S4. Predict the feedforward control quantity and construct the inner PID model based on the output of the outer PID compensation model and the feedforward control quantity. S5. Obtain the adaptation coefficient based on the historical target speed and the output of the inner PID model; S6. Adjust the output of the current inner-layer PID model according to the adaptation coefficient to obtain the speed control quantity.

2. The adaptive speed control method for trains according to claim 1, characterized in that, The input to the outer PID compensation model in S3 is the outer loop error, which is equal to the difference between the target speed and the actual speed. The outer-layer PID compensation model includes: the proportional term of the outer loop error, the integral term of the outer loop error, the differential term of the outer loop error, the slow-time compensation term, and the short-time compensation term.

3. The adaptive speed control method for trains according to claim 1, characterized in that, S4 includes the following sub-steps: S41. Based on the dynamic model, predict the feedforward control quantity: S42. Add the feedforward control quantity to the output of the outer PID compensation model and subtract the actual speed to obtain the feedforward fusion inner loop error. S43. Set the proportional term, piecewise integral term, and band-limited derivative term for the feedforward fusion inner loop error to obtain the inner layer PID model.

4. The adaptive speed control method for trains according to claim 3, characterized in that, The expression for the inner PID model in S43 is: ， Among them, u 2,t e represents the output of the inner PID model at time t. 2,t Let K be the feedforward fusion inner loop error at time t. p2 K is the proportionality coefficient of the inner loop error. i2 K represents the integral coefficient of the inner loop error. d2 I is the differential coefficient of the inner loop error. seg,t For the piecewise integral at time t, D(e 2,t Let be the band-limited differential at time t.

5. The adaptive speed control method for trains according to claim 4, characterized in that, Piecewise integral I seg,t In the feedforward fusion inner loop error e 2,t If the direction of the slow-time compensation is consistent with that of the slow-time compensation, and the absolute value of the slow-time compensation is less than the allowable threshold for the slow-time compensation, then normal integration occurs. Otherwise, an attenuation factor is set, and the attenuation factor is compared with the piecewise integral I of the previous time step. seg,t-1 Multiply, and use the result of the multiplication as the current piecewise integral I. seg,t ; The band-limited differential D(e) at time t 2,t The calculation process includes: at the current time t, obtaining the feedforward fusion inner loop error e. 2,t and the feedforward fusion inner loop error e from the previous time step 2,t-1 The rate of change of error is obtained, and the band-limited differential D(e) of the previous time step is read. 2,t-1 Based on the filtering time constant and sampling period, calculate the filtering coefficients, and then apply the band-limited differential D(e^(-1 / 2)) to the sampled area. 2,t-1 The band-limited differential D(e^t) at time t is obtained by proportionally integrating the error rate of change with the error rate. 2,t ).

6. The adaptive speed control method for trains according to claim 1, characterized in that, S5 includes the following steps: S51. Calculate the ratio of target velocities at adjacent time points to obtain the target velocity change ratio; S52. Calculate the output ratio of the inner-layer PID model at adjacent time points to obtain the control quantity change ratio; S53. Calculate the adaptation coefficient at time t based on the ratio of the change in target velocity to the ratio of the change in control quantity.

7. The adaptive speed control method for trains according to claim 6, characterized in that, The formula for calculating the fitness coefficient at time t in S53 is as follows: ， Where, ζ t Let be the fitness coefficient at time t, exp be the exponential function, and r be the coefficient of fitness. v,t r is the ratio of the target velocity change at time t. u,t Let be the change ratio of the control quantity at time t, ln be the logarithmic function, k be the trend sensitivity coefficient, and || be the absolute value operation.

8. The adaptive speed control method for trains according to claim 1, characterized in that, The expression for the speed control quantity obtained in S6 is: ， Among them, u 2,t,z Let ζ be the velocity control variable at time t. t Let u be the fitness coefficient at time t. 2,t u is the output of the inner PID model at time t. 2,t-1 This represents the output of the inner PID model at time t-1, where t is the time number.

Images