A Constant Speed ​​Control Method and System for Trains Based on Particle Swarm Optimization Algorithm

Abstract

This invention discloses a constant speed control method and system for trains based on particle swarm optimization (PSO) algorithm. The invention relates to the technical field of constant speed control. It determines corresponding PID parameter combinations based on the identification of each operating node, marking the PID parameter combinations of each node. Based on each PID parameter combination, the actual driving environment of the train, and the PSO algorithm, PID parameter optimization events are determined, and each PID parameter combination is further optimized to output the optimal PID parameter combination, thus improving the accuracy of the optimal PID parameter combination. The constant speed state of the train is determined based on the constant speed of each constant speed node and the corresponding constant speed event; the constant speed state of the train is marked; the corresponding constant speed parameter is determined based on the optimal PID parameter combination and the train's constant speed state; and the corresponding constant speed control system is determined based on this constant speed parameter, the actual driving speed of the train, and the corresponding operating state, thus improving the accuracy of the constant speed control system.

Description

Technical Field

[0001] This invention relates to the technical field of constant speed control, and in particular to a constant speed control method and system for trains based on particle swarm optimization algorithm. Background Technology

[0002] With the development of technology, trains are gradually being applied to people's lives and used to carry goods or people. Trains travel in different scenarios and have corresponding driving modes. In the current technology, the train's driving status is monitored in real time, and multiple driving data of the train are collected. The driving status of the train is determined based on multiple driving data, and the corresponding PID parameter combination is marked. At this time, the PID parameter combination is not the optimal PID parameter combination, which affects the accuracy of the constant speed parameter and leads to a decrease in the accuracy of the constant speed control system. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a constant speed control method and system for trains based on particle swarm optimization algorithm.

[0004] This invention provides a constant speed control method for trains based on particle swarm optimization algorithm, comprising: Based on the dynamic detection of the train, multiple driving data of the train at different time periods are determined, and multiple working nodes of the train in the driving phase are determined based on the multiple driving data and the working status of the train. The corresponding PID parameter combination is determined based on the identification of each working node, so as to mark the PID parameter combination of each node. The PID parameter optimization event is determined based on the various PID parameter combinations, the actual driving environment of the train, and the particle swarm optimization algorithm. The various PID parameter combinations are further optimized to output the best PID parameter combination. When a PID parameter optimization event is triggered, it means that the globally optimal PID parameter combination has been found under the dynamically changing objective. The corresponding constant speed range is determined by identifying multiple driving data, and multiple constant speed nodes are determined based on the detection of the constant speed range. The constant speed state of the train is determined based on the constant speed of each constant speed node and the corresponding constant speed event. The constant speed of the constant speed node serves as a representative speed value that can resist instantaneous disturbances. The constant speed state of the train is marked, and the corresponding constant speed parameters are determined based on the optimal combination of PID parameters and the constant speed state of the train. Based on the constant speed parameters, the actual speed of the train and the corresponding operating state, the corresponding constant speed control system is determined.

[0005] This invention provides a constant speed control system for trains based on particle swarm optimization algorithm, which is applied to the aforementioned constant speed control method for trains based on particle swarm optimization algorithm.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) Based on the dynamic detection of the train, multiple driving data of the train in different time periods are determined. Based on the multiple driving data and the working status of the train, multiple working nodes of the train in the driving stage are determined. Based on the identification of each working node, the corresponding PID parameter combination is determined to mark the PID parameter combination of each node. Based on each PID parameter combination, the actual driving environment of the train and the particle swarm optimization algorithm, the PID parameter optimization event is determined, and each PID parameter combination is further optimized to output the best PID parameter combination. Each working node is introduced and each PID parameter combination is further controlled. It takes into account the consideration of each PID parameter combination, the actual driving environment of the train and the particle swarm optimization algorithm, and improves the accuracy of the best PID parameter combination.

[0007] (2) The corresponding constant speed range is determined based on the identification of multiple driving data, and multiple constant speed nodes are determined based on the detection of the constant speed range. The constant speed state of the train is determined based on the constant speed of each constant speed node and the corresponding constant speed event. The constant speed state of the train is marked, and the corresponding constant speed parameter is determined based on the optimal combination of PID parameters and the constant speed state of the train. Based on the constant speed parameter, the actual driving speed of the train and the corresponding working state, the corresponding constant speed control system is determined, and the constant speed state of the train is further controlled. The overall consideration of the constant speed parameter, the actual driving speed of the train and the corresponding working state is realized, and the accuracy of the constant speed control system is improved. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 in the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 in the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 5 This is a flowchart illustrating step S14 in the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 in the constant speed control method for trains based on particle swarm optimization algorithm in an embodiment of the present invention. Figure 7 This is a schematic diagram of the structural composition of a constant speed control system for a train based on particle swarm optimization algorithm in an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 7 A constant speed control method for trains based on particle swarm optimization algorithm is proposed and applied to constant speed control scenarios. The constant speed control method for trains based on particle swarm optimization algorithm includes: Step S11: Based on the dynamic detection of the train, determine multiple travel data of the train at different time periods, and determine multiple working nodes of the train in the travel phase based on the multiple travel data and the working status of the train. Step S12: Determine the corresponding PID parameter combination based on the identification of each working node, so as to mark the PID parameter combination of each node; Step S13: Determine the PID parameter optimization event based on each PID parameter combination, the actual driving environment of the train, and the particle swarm optimization algorithm, and further optimize each PID parameter combination to output the best PID parameter combination. Step S14: Determine the corresponding constant speed range based on the identification of multiple driving data, and determine multiple constant speed nodes based on the detection of the constant speed range. Determine the constant speed state of the train based on the constant speed of each constant speed node and the corresponding constant speed event. Step S15: Mark the constant speed state of the train, determine the corresponding constant speed parameters based on the optimal combination of PID parameters and the constant speed state of the train, and determine the corresponding constant speed control system based on the constant speed parameters, the actual speed of the train and the corresponding working state.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: Mark the current position of the train, determine the corresponding dynamic detection method based on the current position and corresponding speed of the train, and trigger the dynamic detection of the train along the dynamic detection method to determine multiple travel data of the train in different time periods; S112: Collect the train's database, determine the corresponding driving control space based on the detection of the train's database, determine the train's working state based on the traversal of the driving control space, determine the corresponding multimodal data based on multiple driving data, the train's working state and the corresponding driving route, and determine multiple working nodes of the train in the driving phase based on the identification of the multimodal data.

[0012] In the embodiments of this application, the current position of the train is marked, and a corresponding dynamic detection method is determined based on the current position and the corresponding speed of the train. The dynamic detection of the train is triggered along the dynamic detection method to determine multiple driving data of the train in different time periods. This takes into account the overall consideration of the current position and the corresponding speed of the train, and ensures the accuracy of the corresponding dynamic detection method.

[0013] At this point, the system integrates dead reckoning data from the Global Navigation Satellite System, orbital circuits or transponders, inertial measurement units, and velocity sensors, and performs weighted fusion using algorithms such as Kalman filtering. GNSS provides global absolute coordinates, transponders perform error correction at key nodes, and IMU and velocity sensors ensure positioning continuity in areas where GNSS signals are blocked, such as tunnels. The final output is a continuous spatiotemporal position information with confidence assessment, such as K123+456, 116°E…, 39°N…, with a positioning error of <0.5m.

[0014] The system maps the train's spatiotemporal state to specific detection strategies. The system inputs include high-precision position, real-time speed, and track attributes (such as gradient, radius of curvature, tunnels, etc.) retrieved from the track database. Decisions are made through rule bases or machine learning models, such as activating high-speed cruise mode on straight tracks and switching to heavy-load traction mode on steep slopes. The output is a structured configuration file that defines parameters such as sampling frequency, data channels, trigger conditions, and filter settings in detail.

[0015] The system sends commands to various sensors and ECUs through the vehicle network and dynamically adjusts the working mode according to the configuration file; it establishes a real-time data bus and subscribes to different data channels as needed, such as high-frequency acquisition of lateral acceleration data in curve passage mode; before data recording, it performs preprocessing such as timestamp alignment, unit conversion and outlier removal, and finally outputs a highly synchronous multivariate time series dataset that accurately reflects the state of the train at specific time and space points and under specific operating conditions.

[0016] Specifically, when the train enters a 20-kilometer straight section, the fusion positioning system confirms that it is located at K350+000 with a speed of 298 km / h. The decision model immediately matches the high-speed constant speed cruise mode, configures the core control variable sampling frequency to 200Hz, activates channels such as the main converter and brake control unit, and sets speed fluctuations exceeding ±2 km / h as an abnormal trigger condition. The onboard data acquisition system executes accordingly, recording at high frequency the subtle control adjustments required for the train to maintain a constant speed of 300 km / h, providing data support for analyzing the performance of the PID controller under ideal operating conditions.

[0017] When the train approaches a curve with a radius of 4000 meters at 301 km / h, the system recognizes the approaching curve condition at K369+800 and automatically switches to high-speed curve-passing mode. The new strategy increases the sampling frequency of channels such as lateral acceleration to 500Hz, activates the secondary suspension system sensors, and sets a lateral acceleration exceeding 0.6 m / s² as a high-level trigger condition. During the curve-passing process, the system not only records the fine-tuning of traction / braking force but also captures high-frequency changes in vehicle attitude and suspension system response, providing key data for evaluating ride comfort under constant speed control. Through this context-aware adaptive acquisition, the data obtained by the train under different operating conditions has strong relevance and information density, laying a solid foundation for subsequent node identification and parameter optimization.

[0018] Furthermore, a train database is collected, and the corresponding driving control space is determined based on the detection of the train database. The working state of the train is determined based on the traversal of the driving control space. Based on multiple driving data, the working state of the train and the corresponding driving route, the corresponding multimodal data is determined. Based on the identification of the multimodal data, multiple working nodes of the train in the driving phase are determined. This incorporates the overall consideration of multimodal data identification and ensures the accuracy of multiple working nodes of the train in the driving phase.

[0019] At this point, the system integrates the train's static design parameters (such as mass and traction / braking characteristic curves), dynamic state data (such as current speed and load), and environmental information (such as track gradient and curvature). Through the construction method of multi-dimensional constraint space, it calculates the feasible domain of traction and braking force under the current operating conditions. This feasible domain is subject to multiple constraints such as physical limits, wheel-rail adhesion, safety margin, and ride comfort. Finally, it outputs a dynamic model that clearly defines the range of safe and smooth control actions.

[0020] The system uses pattern recognition algorithms to analyze the position, trend of change, and relative relationship with the boundary of the actual control commands in the control space, thereby semantically classifying the current working state. For example, if the traction command is far below the upper limit and changes gently, it is identified as smooth cruise; if the command continues to approach the upper limit but the acceleration is still decreasing, it is determined to be heavy-load traction on a steep slope. Finally, it outputs one or more clearly defined working state labels.

[0021] The system precisely aligns and associates the original driving data stream of S111, the semantic working status labels output by sub-step 2, and the driving route information from the route database with timestamps; thereby constructing a structured multimodal dataset, in which each record not only contains numerical values, but also has semantic labels that describe its background, forming a continuous data stream that is rich in information and has a clear context.

[0022] The algorithm employs rule-based state machine or time series analysis to identify key operational phases with clear start and end markers. For example, when the constant speed cruise state continues for more than a preset time, it is marked as a constant speed cruise node. The transition point of the state is defined as the start or end of the node. Finally, a discrete sequence of working nodes arranged in chronological order is output, and each node contains complete information such as time, state, and key statistical data.

[0023] Specifically, when the train enters a downhill slope of -10‰ at a speed of 300km / h, the system constructs its driving control space. By integrating the train's full load mass, the regenerative braking characteristic curve at 300km / h, and the additional force generated by the -10‰ slope, the system calculates that in order to maintain a constant speed, a regenerative braking force of 50kN to 200kN must be applied. This range precisely defines the safety control boundary under the current operating conditions.

[0024] The system compares the actual braking command (stabilized at around 75kN) collected by S111 with the control space. Since 75kN is near the lower limit and fluctuates very little, the system intelligently identifies the current working state as downhill constant speed regenerative braking. At this moment, the driving data, working state and route information are merged into a multimodal record, such as {timestamp: 10:30:15, speed: 300.1km / h, braking force: 75kN, working state: downhill constant speed regenerative braking, driving route: -10‰ downhill K520+300}.

[0025] For the next 5 minutes, as long as this state persists, the system will identify this continuous stable operating condition through the rule engine. When the train finally leaves the downhill slope and the operating state changes to flat constant speed traction, the system accurately captures this state transition point and completely divides and marks this 5-minute operation process as an independent downhill constant speed cruise node. This node will become a valuable data basis for subsequent optimization of PID parameters for specific operating conditions.

[0026] refer to Figure 3 In step S12, the specific steps are as follows: S121: In multiple working nodes, the corresponding data set is determined based on the tracing of each working node, and multiple PID parameters are determined based on the identification of the data set. The corresponding combination of PID parameters is determined based on the multiple PID parameters, the train's driving position and the corresponding working state. S122: Collect each node and its corresponding PID parameter combination, determine the corresponding matching coefficient based on the matching between the node and the PID parameter combination, and if the matching coefficient is lower than the preset matching coefficient threshold, trigger the reverse tracing of the node until the matching coefficient between the PID parameter combination and the node is greater than the preset matching coefficient threshold.

[0027] In the embodiments of this application, among multiple working nodes, a corresponding data set is determined based on the tracing of each working node, and multiple PID parameters are determined based on the identification of the data set. A corresponding PID parameter combination is determined based on the multiple PID parameters, the train's driving position, and the corresponding working state, which takes into account the overall consideration of multiple PID parameters, the train's driving position, and the corresponding working state, and ensures the accuracy of the corresponding PID parameter combination.

[0028] At this point, the system abstracts the current working node into a feature vector containing information such as working status, target speed, line gradient, and load level. Using this vector, it performs an approximate nearest neighbor search or multi-condition filtering query in the historical database, quantifies the similarity between the current node and historical cases through a defined distance function, and finally outputs a data set consisting of highly similar historical running segments, laying the foundation for subsequent analysis.

[0029] The system can adopt two technical approaches: one is direct parameter extraction, which directly reads the PID parameter values ​​that are in effect at that time from the metadata attached to the historical data; the other is system identification and parameter estimation, which applies system identification algorithms such as recursive least squares when the historical data only contains process variables, treating each historical data segment as a black box test, and back-estimating the PID controller model parameters that can reproduce its input-output relationship; finally, a list containing multiple candidate PID parameter values ​​is output.

[0030] The system clusters the multiple candidate parameters obtained in the previous step in the three-dimensional parameter space [Kp, Ki, Kd], so that parameters with similar physical meanings are naturally grouped together. For each cluster, the system calculates its center point or selects the actual data point closest to the center point as the representative PID parameter combination of the cluster. At the same time, the system analyzes the subtle differences in driving position and working state of the historical cases corresponding to each parameter combination, and assigns them more refined labels. Finally, it outputs one or more highly reliable PID parameter combinations that have been summarized and labeled.

[0031] Specifically, when the train identifies the operating node of constant speed cruising at 300 km / h on a -10‰ gradient, S121 begins to match initial parameters for it; the system characterizes this node as {state: constant speed, speed: 300, gradient: -10, load: full load} and initiates a precise query in the historical database; the system successfully retrieves 8 highly matching historical operation segments, the target speed of these segments is between 295 and 305 km / h, the gradient is between -12‰ and -8‰, and all are in a full load state, which together constitute the data set for this analysis.

[0032] The system checked the metadata of these 8 historical segments and found that they all fully recorded the PID parameters used at that time. Therefore, the direct parameter extraction method was adopted. The system quickly extracted 8 sets of specific PID parameter values, such as [Kp:1.2,Ki:0.05,Kd:0.8], [Kp:1.25,Ki:0.048,Kd:0.82], etc., forming a candidate parameter list.

[0033] The system inputs these 8 sets of parameters into the K-Means clustering algorithm for analysis. The algorithm divides the parameters into two clusters: cluster A contains 5 sets of parameters with relatively low Kd values, and cluster B contains 3 sets of parameters with relatively high Kd values. Through further correlation analysis, the system found that the 3 historical segments corresponding to cluster B all occurred under weather conditions with crosswinds or light rain. Based on this, the system calculates the centroids of the two clusters and finally generates two representative PID parameter combinations with clear applicable scenarios: one is a standard downhill constant speed combination suitable for sunny weather, and the other is a disturbed downhill constant speed combination suitable for disturbed environments. This provides a high-quality starting point for subsequent parameter verification and optimization.

[0034] Furthermore, each node and its corresponding PID parameter combination are collected. The matching coefficient is determined based on the matching between the node and the PID parameter combination. If the matching coefficient is lower than the preset matching coefficient threshold, the node is triggered to trace back until the matching coefficient between the PID parameter combination and the node is greater than the preset matching coefficient threshold. This takes into account the overall matching of the node and the PID parameter combination and ensures the accuracy of the corresponding matching coefficient.

[0035] At this point, the system will conduct a comprehensive evaluation from multiple sub-indicators, including the similarity of working condition features obtained by calculating the distance between feature vectors, the historical performance score obtained by re-evaluating the control performance in the simulation model, the confidence of the data source based on the quality and quantity of historical data, and the parameter universality reflected by the degree of parameter dispersion. These four sub-indicators are weighted and fused by preset weights, and finally output a matching coefficient between 0 and 1. The higher the value, the more reliable the matching.

[0036] The system sets a hard matching coefficient threshold as the minimum acceptable standard. When the evaluated matching coefficient is lower than this threshold, the system will reject the current match and immediately trigger reverse tracing. This tracing is not a simple repetition, but an intelligent backtracking process that gradually relaxes the search conditions. For example, it first relaxes non-core constraints such as slope or speed. If it still does not meet the standard, it further relaxes core constraints such as load or working status. Each time a backtracking is executed, the system will re-perform parameter matching and coefficient evaluation. This cycle will continue until a parameter combination with a matching coefficient exceeding the threshold is found, or the maximum backtracking level is reached and conservative default parameters are enabled.

[0037] Specifically, assuming the train in S121 is matched with an initial PID parameter combination A for the working node of R4500m curve, 300km / h, moderate crosswind, and constant speed cruise; S122 then begins to verify its quality; the system collects the characteristics of the current node and the source information of combination A, and evaluates it from four dimensions: since combination A comes from clear weather data, it differs from the current working condition with crosswind, and the working condition similarity score is low; in the simulation, the combination has a large speed fluctuation due to not considering crosswind disturbance, and the historical performance score is poor; however, its source data is sufficient, and the data confidence and parameter universality scores are high; after weighted calculation, the final matching coefficient is 0.7825.

[0038] Since the coefficient is lower than the preset threshold of 0.85, the system determines that combination A is unreliable and immediately triggers the reverse tracing mechanism. In the first-level rollback, the system relaxes the weather condition restrictions, re-searches the historical database for constant velocity curve data containing any wind disturbances, and generates a new PID parameter combination B based on the three new data segments found. The system evaluates combination B again. Since the new data also contains crosswinds, the operating condition similarity and historical performance scores are significantly improved. After recalculation, the new matching coefficient reaches 0.91, successfully exceeding the threshold. The system accepts combination B as the initial PID parameter combination for the current node, ends the S12 process, and prepares for the fine-tuning optimization in S13.

[0039] refer to Figure 4 In step S13, the specific steps are as follows: S131: Determine the train model based on the detection of the train's database, determine the particle swarm optimization algorithm based on the train model, the speed of the type of train and the corresponding current position, and determine the corresponding optimization system based on the identification of the particle swarm optimization algorithm; S132: Determine the first level of optimization content based on the optimization system and the combination of various PID parameters. At the same time, collect the actual driving environment of the train and determine the second level of optimization content based on the optimization system and the actual driving environment of the train. Determine the PID parameter optimization event based on the multiple iterations of the first and second level of optimization content. S133: Based on the identification of the PID parameter optimization event, multiple parameter optimization projects are determined. The optimization process of the PID parameter combination is determined according to the project content, corresponding project priority and each PID parameter combination. The optimal PID parameter combination is determined based on the identification of the optimization process of the PID parameter combination.

[0040] In the embodiments of this application, the train model is determined based on the detection of the train database, the particle swarm optimization algorithm is determined based on the train model, the speed of the type of train and the corresponding current position, and the corresponding optimization system is determined based on the identification of the particle swarm optimization algorithm. This approach is compatible with the overall consideration of the identification of the particle swarm optimization algorithm and ensures the accuracy of the corresponding optimization system.

[0041] At this point, the system accesses the train configuration database and reads the unique identifier in the control network to accurately identify the specific model of the train. Once the model is determined, the system automatically loads a complete set of associated dynamic parameters, including total mass, moment of inertia, traction / braking characteristic curves, and running resistance formulas, providing a data foundation for the subsequent construction of a high-fidelity simulation environment.

[0042] The system has a built-in algorithm configurator that comprehensively analyzes the complexity of the controlled object determined by the train model, the dynamic response characteristics affected by the current speed, and the track environment information provided by the current location. It then intelligently adjusts the core components of the PSO algorithm, including setting the population size based on dynamic complexity and track conditions, selecting the inertial weight strategy based on the degree of dynamic change in the operating scenario, and dynamically adjusting the learning factor based on the magnitude of environmental disturbances, ultimately achieving a precise match between the algorithm and the task.

[0043] The system integrates the configured PSO algorithm core with other necessary components to form an executable optimization system. This system not only includes the PSO algorithm, but also integrates a high-fidelity simulation environment built based on train parameters and track information, a multi-objective fitness function preset according to train type and scenario, a constraint processing module to ensure the feasibility of parameter engineering, and a parallel framework that uses multi-core or GPU resources to accelerate computation. Finally, it outputs a fully customized optimization solution that can run automatically.

[0044] Specifically, when the train (model CR400AF-B intelligent EMU) is about to enter a complex mountainous section with long tunnels and continuous small-radius curves at a speed of 350km / h, S131 begins to tailor its configuration; the system accurately identifies its model through the train database and then loads the unique permanent magnet traction system characteristics, lightweight car body drag coefficient, and precision suspension system parameters of that model.

[0045] The system comprehensively analyzes the complex dynamics of the CR400AF-B, its extremely high speed of 350km / h, and the variable environment of complex mountain road sections, and intelligently configures the PSO algorithm. The algorithm configurator decides to use a larger population size of 200 to cope with nonlinearity, selects adaptive chaotic perturbation inertial weights to avoid getting trapped in local optima, and increases the individual learning factor to encourage self-exploration in uncertain environments. At the same time, it configures stricter constraint processing to ensure comfort at high speeds.

[0046] The system integrates all components into a dedicated optimization system; the simulation environment of this system dynamically loads accurate route data of the preceding road section and simulates tunnel air pressure and curve centrifugal force effects; the fitness function, due to its high-end passenger train nature, sets the comfort weight to the highest level, reflecting the concept of prioritizing comfort; at the same time, GPU parallel computing is activated to accelerate the simulation evaluation of 200 particles; through this series of refined configurations, S131 has created a powerful and highly customized optimization system for the constant speed control optimization task of the train under extreme conditions.

[0047] Furthermore, the first level of optimization content is determined based on the optimization system and the combination of various PID parameters. At the same time, the actual driving environment of the train is collected, and the second level of optimization content is determined based on the optimization system and the actual driving environment of the train. The PID parameter optimization event is determined based on the multiple iterations of the first and second level of optimization content, which is compatible with the overall consideration of the multiple iterations of the first and second level of optimization content, and ensures the accuracy of the PID parameter optimization event.

[0048] At this point, the system generates an initial particle swarm in the three-dimensional [Kp, Ki, Kd] space, centered on the parameter combination provided by S12. In each iteration, each particle representing a set of candidate parameters is substituted into the high-fidelity simulation environment and its fitness value is calculated based on its output response. The PSO algorithm updates the individual and global optimal positions based on the fitness values ​​of all particles and generates a new generation of particle swarm, driving the entire swarm to gradually approach the optimal parameter region.

[0049] The system collects key variables such as line geometry, environmental disturbances, and operational status in real time through onboard sensors and a line database. The built-in environment-weight mapping model dynamically adjusts the weights of each performance index in the fitness function based on this real-time data. For example, it increases the weight of comfort when entering a small radius curve and strengthens the weight of robustness-related indicators when encountering strong crosswinds, so that the optimization target itself can adapt to environmental changes in real time.

[0050] Before each iteration of PSO begins, the system updates the fitness function weights based on environmental information and uses this new objective function to evaluate and update the current particle swarm. This cycle of perceiving the environment, adjusting the objective, and optimizing parameters continues until the termination condition of PSO is met. At this point, a PID parameter optimization event is triggered, indicating that the globally optimal combination of PID parameters has been found under the dynamically changing objective.

[0051] Specifically, when the train is about to enter a complex section of track containing tunnels, strong crosswinds, and continuous curves at a speed of 300 km / h, the S132's two-layer collaborative optimization mechanism starts to operate; the first optimization, namely the PSO algorithm, starts with the initial parameter combination, begins to explore within the parameter space, and uses a standard fitness function to perform a preliminary evaluation of the particles.

[0052] Meanwhile, the second level of optimization responds to environmental changes in real time. When the system detects that the train is entering a tunnel, it immediately adjusts the fitness function and slightly increases the steady-state accuracy weight to cope with changes in air resistance inside the tunnel. When the train is about to exit the tunnel, it predicts sudden changes in air pressure and crosswinds and significantly increases the impact weight to prepare for disturbance resistance. When the crosswind sensor detects a strong wind of 8 m / s, the system further increases the robustness-related weight and sets disturbance resistance as the highest priority. After entering a continuous curve, it maintains a high comfort weight to ensure smooth cornering.

[0053] Within minutes of the train passing through the section, the two optimization processes underwent hundreds of coordinated iterations; the search direction of the particle swarm evolved continuously with the dynamic adjustment of the fitness function: from the initial steady-state control within the tunnel, to the mid-term robust control against impacts and strong crosswinds upon exiting the tunnel, and then to the later cornering comfort control; when the maximum number of iterations was reached, the PID parameter optimization event was triggered; the optimal PID parameter combination output by the system is a parameter set with the best overall performance after undergoing various dynamic environments, and its robustness and adaptability far exceed the results optimized under any single working condition.

[0054] Therefore, based on the identification of the PID parameter optimization event, multiple parameter optimization projects are determined. The optimization process of each PID parameter combination is determined according to its content, priority, and the specific PID parameter combinations. The optimal PID parameter combination is then determined based on the identification of this optimization process. This approach considers the overall process of identifying the optimization process of PID parameter combinations, ensuring the accuracy of the optimal PID parameter combination. Furthermore, various working nodes are introduced, and each PID parameter combination is further controlled. This approach considers various PID parameter combinations, the actual operating environment of the train, and the particle swarm optimization algorithm, thereby improving the accuracy of the optimal PID parameter combination.

[0055] At this point, based on the engineering requirements of constant speed control of the train, the system defines a series of standardized parameter optimization projects, focusing on key performance indicators such as basic stability, dynamic response performance, energy consumption economy, ride comfort, and robustness and disturbance rejection. This multi-objective decomposition method ensures that the optimization process can comprehensively cover all aspects of control performance.

[0056] The system dynamically assigns priorities to each project based on the train model, operating route, and current operating conditions. The system executes these optimization projects in descending order of priority, and the output of each stage becomes the input and constraint of the next stage. For example, the system first finds the parameters with the best comfort, and based on this, it further searches for the parameters with the best stability while satisfying comfort constraints. Through this meticulous approach, the system gradually approaches the global optimum, ultimately forming a complete optimization process trajectory.

[0057] The system backtracks the entire optimization process and performs a final evaluation of the candidate optimal parameter combinations generated at the end of each stage. The evaluation uses a comprehensive fitness function with fixed weights that includes all performance indicators, and the weights reflect the top-level design requirements of the system. The system selects the highest-scoring combination of all candidate parameter combinations as the optimal PID parameter combination by comparing the comprehensive scores of all candidate parameter combinations. This decision-making process ensures that the final parameters do not excel in a single objective.

[0058] Specifically, after the train completes the dynamic optimization of S132 and triggers the PID parameter optimization event, S133 begins the final stage and decision-making. The system defines five parameter optimization items for the train: P1 (comfort), P2 (stability), P3 (dynamic performance), P4 (energy consumption), and P5 (robustness). Considering that the train is a high-end passenger train, the system assigns it a priority of P1>P2>P3>P4>P5.

[0059] The system begins to build an optimization process; in the first stage, the PSO algorithm runs and finds a set of parameters Para_1 that minimizes the impact; in the second stage, the algorithm performs a fine search centered on Para_1, and finds the more stable parameter Para_2 while keeping comfort almost unchanged; subsequent stages are carried out in sequence, and the system continuously optimizes dynamic performance, energy consumption and robustness based on the previous optimization, and finally forms a complete optimization process from Para_1 to Para_5.

[0060] The system makes the final decision; it traces back the entire process, extracts candidate parameter combinations for five stages, and uses a final comprehensive fitness function (such as comfort weight 0.3, stability weight 0.3, etc.) to score all candidate parameters. The evaluation results show that although Para_1 scores the highest in comfort, its scores in other items are relatively low; while Para_5, although not the highest in any single item, has the most balanced indicators, no obvious weaknesses, and the highest overall score. Therefore, the system determines that Para_5 is the best PID parameter combination for the train under this operating condition. This is the most balanced and reliable solution obtained after phased, multi-objective, and full-process optimization.

[0061] refer to Figure 5 In step S14, the specific steps are as follows: S141: Collect multiple driving data of the train, determine the dynamic driving curve of the train based on the multiple driving data and the actual driving environment of the train, determine multiple dynamic driving nodes based on the detection of the dynamic driving curve, and determine the corresponding constant speed range based on the node position of each dynamic driving node, the corresponding speed range and the train's previous constant speed events. S142: Based on the matching of the constant speed range and the dynamic travel curve of the train, multiple constant speed nodes are determined, and the constant speed of each constant speed node is marked. The constant speed state of the train is determined based on the node position, the corresponding constant speed, and the corresponding constant speed event of each constant speed node. In the embodiments of this application, multiple train driving data are collected, and a dynamic driving curve of the train is determined based on the multiple driving data and the actual driving environment of the train. Multiple dynamic driving nodes are determined based on the detection of the dynamic driving curve. The corresponding constant speed range is determined based on the node position of each dynamic driving node, the corresponding speed range, and the train's previous constant speed events. This approach takes into account the overall consideration of the node position of each dynamic driving node, the corresponding speed range, and the train's previous constant speed events, ensuring the accuracy of the corresponding constant speed range.

[0062] At this point, the system will align the core driving data, such as high-frequency speed and acceleration, from the onboard sensors with environmental information such as gradient and curvature from the track database in time and space. By using state estimation algorithms such as Kalman filters, the system will fuse noisy measurement data and physical constraints to effectively suppress noise and estimate the accurate state. Finally, it will output a smooth, multi-dimensional dynamic driving curve that fully reflects the real motion state of the train at different time and space points.

[0063] The system analyzes the velocity and acceleration curves in the dynamic driving curve and uses algorithms such as thresholding or curvature methods to capture the instant of state switching. For example, when the acceleration fluctuates continuously in the range close to zero, it is marked as a uniform speed state; when it changes from a positive value to zero or from zero to a negative value, it is marked as a dynamic driving node. Finally, it outputs a node sequence arranged in chronological order that describes the macroscopic driving mode switching of the train.

[0064] For each identified constant speed segment, the system extracts key indicators such as its duration, speed fluctuation range, and mileage span. These indicators are then compared with a constant speed event model based on historical data statistics. This model defines necessary conditions for effective constant speed control, such as the minimum duration, maximum speed fluctuation threshold, and minimum mileage. Only constant speed segments that fully meet all conditions are formally identified as an effective constant speed range.

[0065] Specifically, after the train completes a run, S141 begins to sift through massive amounts of data to select meaningful constant-speed segments. The system collects the train's raw speed data at 100Hz and the track's gradient data. Through a Kalman filter, the system fuses these data, filters out high-frequency noise from the sensors, and accurately compensates for the impact of gradient changes on acceleration, ultimately generating a smooth and accurate dynamic travel curve, laying the foundation for subsequent analysis.

[0066] The system analyzed the acceleration changes on the curve and successfully identified a series of dynamic driving nodes. For example, the system found that at t=500 seconds, the acceleration dropped from a positive value to 0, marking it as the beginning node of uniform speed; at t=1400 seconds, the acceleration changed from 0 to a negative value, marking it as the end node of uniform speed. Thus, the system locked a uniform speed segment from 500 seconds to 1400 seconds.

[0067] The system rigorously screens this segment; it calculates the segment's duration to be 900 seconds (15 minutes), its speed fluctuation range to be 2.7 km / h, and its mileage span to be over 80 kilometers. After comparing these indicators with the thresholds of the constant speed event model, the system confirms that the segment fully meets all conditions, including minimum duration, maximum speed fluctuation, and minimum mileage. Therefore, the segment is officially determined to be a valid constant speed range, and its location and speed information will be transmitted to S142 for the next step of precise node definition.

[0068] Furthermore, multiple constant speed nodes are determined based on the matching between the constant speed range and the train's dynamic travel curve, and the constant speed of each constant speed node is marked. The constant speed state of the train is determined based on the node position, corresponding constant speed, and corresponding constant speed event of each constant speed node. This approach takes into account the overall consideration of the node position, corresponding constant speed, and corresponding constant speed event of each constant speed node, ensuring the accuracy of the train's constant speed state.

[0069] At this point, the system accurately maps the start and end mileage markers of the constant speed range back onto the high-fidelity dynamic driving curve, finds the corresponding precise timestamps through interpolation and search, and thus anchors the node in time and space. Within the anchored time range, the system extracts all speed sampling points and uses robust statistical methods such as arithmetic mean or median to calculate the constant speed of the constant speed node. The constant speed of the constant speed node serves as a representative speed value that can resist instantaneous disturbances.

[0070] The system collects the location information, constant speed, and start and end methods (i.e., constant speed events) of each node; it then matches this contextual information with a predefined constant speed state semantic library using a rule engine or decision tree; for example, a node located on a downhill slope and traveling at high speed is marked as downhill high-speed cruise, while a node that enters by slowing down and is located in front of a station is marked as approaching a station speed-limited cruise; each node is assigned a precise and unambiguous state label that is directly associated with a specific control requirement.

[0071] Specifically, the train is identified as having a constant speed range at the mileage marker [K35+100, K115-800]; S142 begins precise anchoring of this range; the system matches the dynamic travel curve to determine the precise timestamp corresponding to this range as 10:15:30 to 10:30:30, thus creating a constant speed node named Node_Cruise_A; then, the system extracts all speed data within this time window and calculates its arithmetic mean, obtaining 300.2 km / h, which is then marked as the constant speed of this node.

[0072] The system enters the semantic classification stage; it collects the context information of the node: its location shows that the first half is a straight track and the second half is a long curve with a radius of 6000 meters; the speed is 300.2 km / h, which falls into the high-speed category; the node enters through acceleration from the start; the system's rule engine matches the straight track and high speed to a straight high-speed cruise state, but then finds that the node contains a long curve, thus triggering a more refined matching rule; the system determines that this is a composite state and assigns it a more precise label: high-speed cruise transitioning from a straight track to a curve; at this point, S142 outputs a highly accurate and semantic constant speed node, providing a direct basis for S15 to match the PID parameter combination most suitable for this transitional condition.

[0073] refer to Figure 6 In step S15, the specific steps are as follows: S151: Real-time monitoring of the constant speed status of the train, and determination of the corresponding constant speed mapping relationship based on the identification of the constant speed status. At the same time, the optimal combination of PID parameters is collected, and the corresponding constant speed parameters are determined according to the optimal combination of PID parameters, the constant speed mapping relationship corresponding to the constant speed status, and the corresponding constant speed event. S152: Collect the actual speed of the train and mark the speed change during the train's operation. Determine the first level of constant speed control based on the actual speed of the train, the corresponding speed change, and the constant speed parameter. S153: Collect the train's operating status, determine the second level of constant speed control content based on the train's actual speed, the corresponding speed change, and the train's operating status, and determine the corresponding constant speed control system based on the first and second level of constant speed control content. In the embodiments of this application, the constant speed state of the train is monitored in real time, and the corresponding constant speed mapping relationship is determined based on the identification of the constant speed state. At the same time, the optimal combination of PID parameters is collected, and the corresponding constant speed parameters are determined according to the optimal combination of PID parameters, the constant speed mapping relationship corresponding to the constant speed state, and the corresponding constant speed event. This approach takes into account the overall consideration of the optimal combination of PID parameters, the constant speed mapping relationship corresponding to the constant speed state, and the corresponding constant speed event, ensuring the accuracy of the corresponding constant speed parameters.

[0074] At this point, the system continuously runs the S14 logic at a high frequency to analyze the real-time data stream and identify the current constant speed state of the train. Simultaneously, the system maintains a core knowledge base, namely the constant speed mapping relationship, which defines the mapping rules from various constant speed states to the corresponding control strategies. This knowledge base is a direct product of the S13 parameter optimization and the S14 state definition. Once the current state is identified, the system immediately queries the mapping relationship to find the entry that matches it perfectly, providing an accurate index for subsequent parameter loading.

[0075] Based on the mapping relationship determined in the previous step, the system loads the corresponding optimal PID parameter combination from the optimization result database of S13 as the basis; the system analyzes the constant speed event that triggers the current state and intelligently adjusts the basic parameters according to the built-in event-parameter fine-tuning rule library, such as appropriately reducing the integral gain when entering a downhill section to prevent integral saturation; to ensure a smooth transition, the system uses a gradient function to make the old and new parameters transition smoothly, and finally outputs a set of fine-tuned and smoothed constant speed parameters that can be directly written to the controller.

[0076] Furthermore, the actual speed of the train is collected, and the speed change during the train's operation is marked. Based on the actual speed of the train, the corresponding speed change, and the constant speed parameter, the first level of constant speed control content is determined. This takes into account the actual speed of the train, the corresponding speed change, and the constant speed parameter as a whole, ensuring the accuracy of the first level of constant speed control content.

[0077] At this point, the system performs multi-source speed measurement using onboard Doppler radar, axle speed sensors, and satellite positioning system, and uses algorithms such as Kalman filtering for data fusion to obtain a high-precision, high-update-rate, and low-noise actual driving speed. Based on this, the system calculates and marks the speed change in real time, i.e., the speed deviation e(t), which is the difference between the target speed and the actual speed. At the same time, to meet the requirements of PID control, the system also calculates and stores the integral and derivative terms of the deviation, providing complete input for subsequent control law calculations.

[0078] The system takes the real-time calculated speed deviation e(t), its historical integral and derivative terms, and the loaded constant speed parameters [Kp_final, Ki_final, Kd_final] provided by S151 as inputs. According to the PID algorithm, the system calculates the outputs of the proportional term (P), integral term (I), and derivative term (D) respectively. The P term directly responds to the current error, the I term accumulates historical errors to eliminate steady-state errors, and the D term predicts error trends to reduce overshoot. The system adds these three outputs together to synthesize a basic control instruction u_base(t), which represents the theoretically applicable traction or braking force calculated by the PID controller based on the feedback error.

[0079] Therefore, by collecting the train's operating status, determining the second level of constant speed control based on the train's actual speed, corresponding speed changes, and operating status, and then determining the corresponding constant speed control system based on the first and second levels of constant speed control, the system integrates the overall considerations of the first and second levels of constant speed control, ensuring the accuracy of the corresponding constant speed control system. At the same time, it further controls the train's constant speed status, realizing the overall consideration of the constant speed parameter, the train's actual speed, and the corresponding operating status, thus improving the accuracy of the constant speed control system. At this point, the system collects more macroscopic train operating conditions than speed, including traction / braking conditions, energy flow status, and system health status. The system then enters a multi-source information fusion decision-making process: on the one hand, based on the known physical model and track data, it calculates a feedforward control quantity that can offset foreseeable disturbances (such as the gradient ahead); on the other hand, it logically corrects the control commands according to the current operating state to avoid conflicts between simultaneous traction and braking operations; at the same time, it prevents sudden changes in commands and ensures the smoothness of the control process by using a command change rate limiter and a smoothing filter.

[0080] The system weighted and fused the first control content output by S152 with the second control content of S153 (including feedforward compensation and operating condition correction). The fusion weight can be dynamically adjusted according to the operating conditions. For example, when the operating conditions are stable, the feedforward weight is increased to improve the response speed, and when the operating conditions change abruptly, the feedback weight is increased to enhance robustness. This complete closed loop from state monitoring to command fusion constitutes a constant speed control system that integrates adaptability and predictability, and finally outputs a fully optimized final control command sent to the train execution unit.

[0081] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of a train constant speed control system based on particle swarm optimization algorithm according to an embodiment of the present invention; the train constant speed control system based on particle swarm optimization algorithm includes: The working node module 21 is used to determine multiple travel data of the train at different time periods based on the dynamic detection of the train, and to determine multiple working nodes of the train in the travel phase based on the multiple travel data and the working status of the train. The PID parameter combination module 22 is used to determine the corresponding PID parameter combination based on the identification of each working node, so as to mark the PID parameter combination of each node. The parameter optimization module 23 is used to determine the PID parameter optimization event based on the various PID parameter combinations, the actual driving environment of the train and the particle swarm optimization algorithm, and further optimize the various PID parameter combinations to output the best PID parameter combination. The constant speed state module 24 is used to determine the corresponding constant speed range based on the identification of multiple driving data, and to determine multiple constant speed nodes based on the detection of the constant speed range, and to determine the constant speed state of the train based on the constant speed of each constant speed node and the corresponding constant speed event. The constant speed control system module 25 is used to mark the constant speed state of the train, determine the corresponding constant speed parameters based on the optimal combination of PID parameters and the constant speed state of the train, and determine the corresponding constant speed control system based on the constant speed parameters, the actual speed of the train and the corresponding working state.

[0082] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A constant speed control method for trains based on particle swarm optimization algorithm, characterized in that, include: Based on the dynamic detection of the train, multiple driving data of the train at different time periods are determined, and multiple working nodes of the train in the driving phase are determined based on the multiple driving data and the working status of the train. The corresponding PID parameter combination is determined based on the identification of each working node, so as to mark the PID parameter combination of each node. The PID parameter optimization event is determined based on the various PID parameter combinations, the actual driving environment of the train, and the particle swarm optimization algorithm. The various PID parameter combinations are further optimized to output the best PID parameter combination. When a PID parameter optimization event is triggered, it means that the globally optimal PID parameter combination has been found under the dynamically changing objective. The corresponding constant speed range is determined by identifying multiple driving data, and multiple constant speed nodes are determined based on the detection of the constant speed range. The constant speed state of the train is determined based on the constant speed of each constant speed node and the corresponding constant speed event. The constant speed of the constant speed node serves as a representative speed value that can resist instantaneous disturbances. The constant speed state of the train is marked, and the corresponding constant speed parameters are determined based on the optimal combination of PID parameters and the constant speed state of the train. Based on the constant speed parameters, the actual speed of the train and the corresponding operating state, the corresponding constant speed control system is determined.

2. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 1, characterized in that, The method involves determining multiple train travel data points at different time periods based on dynamic train detection, and then determining multiple operational nodes of the train during the travel phase based on these multiple travel data points and the train's operational status, including: The current position of the train is marked, and the corresponding dynamic detection method is determined based on the current position and the corresponding speed. The dynamic detection of the train is triggered along the dynamic detection method to determine multiple travel data of the train at different time periods. The system collects a database of train data, determines the corresponding driving control space based on the detection of the train database, determines the train's operating status based on the traversal of the driving control space, determines the corresponding multimodal data based on multiple driving data, the train's operating status and the corresponding driving route, and determines multiple working nodes of the train in the driving phase based on the identification of the multimodal data.

3. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 1, characterized in that, The step of determining the corresponding PID parameter combination based on the identification of each working node, to mark the PID parameter combination of each node, includes: In multiple working nodes, the corresponding data set is determined based on the tracing of each working node, and multiple PID parameters are determined based on the identification of the data set. Based on the multiple PID parameters, the train's driving position and the corresponding working state, the corresponding combination of PID parameters is determined. Collect data from each node and its corresponding PID parameter combination. Determine the matching coefficient based on the match between the node and the PID parameter combination. If the matching coefficient is lower than the preset matching coefficient threshold, trigger the reverse tracing of the node until the matching coefficient between the PID parameter combination and the node is greater than the preset matching coefficient threshold.

4. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 1, characterized in that, The process of determining PID parameter optimization events based on various PID parameter combinations, the actual train operating environment, and the particle swarm optimization algorithm, and further optimizing each PID parameter combination to output the optimal PID parameter combination, includes: The train model is determined based on the detection of the train's database. The particle swarm optimization algorithm is determined based on the train model, the speed of the similar train, and the corresponding current position. The corresponding optimization system is determined based on the identification of the particle swarm optimization algorithm.

5. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 4, characterized in that, The process of determining PID parameter optimization events based on various PID parameter combinations, the actual train operating environment, and the particle swarm optimization algorithm, and further optimizing each PID parameter combination to output the optimal PID parameter combination, also includes: The first optimization content is determined based on the optimization system and the combination of various PID parameters. At the same time, the actual driving environment of the train is collected, and the second optimization content is determined based on the optimization system and the actual driving environment of the train. The PID parameter optimization event is determined based on the multiple iterations of the first and second optimization contents. Based on the identification of the PID parameter optimization event, multiple parameter optimization projects are determined. The optimization process of the PID parameter combination is determined according to the project content, corresponding project priority, and each PID parameter combination. Based on the identification of the optimization process of the PID parameter combination, the optimal PID parameter combination is determined.

6. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 1, characterized in that, The process of determining a corresponding constant speed range based on the identification of multiple driving data, determining multiple constant speed nodes based on the detection of this constant speed range, and determining the constant speed state of the train based on the constant speed of each constant speed node and the corresponding constant speed event includes: Multiple train driving data points are collected, and the train's dynamic driving curve is determined based on these data points and the train's actual driving environment. Multiple dynamic driving nodes are determined based on the detection of these dynamic driving curves, and the corresponding constant speed range is determined based on the node position of each dynamic driving node, the corresponding speed range, and the train's past constant speed events.

7. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 6, characterized in that, The process of determining a corresponding constant speed range based on the identification of multiple driving data, determining multiple constant speed nodes based on the detection of the constant speed range, and determining the constant speed state of the train based on the constant speed of each constant speed node and the corresponding constant speed event further includes: Multiple constant speed nodes are determined based on the matching of the constant speed range and the dynamic travel curve of the train, and the constant speed of each constant speed node is marked. The constant speed state of the train is determined based on the node position, the corresponding constant speed, and the corresponding constant speed event of each constant speed node.

8. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 1, characterized in that, The constant speed state of the marked train is determined by identifying the corresponding constant speed parameters based on the optimal PID parameter combination and the train's constant speed state. Based on these constant speed parameters, the train's actual speed, and the corresponding operating state, a corresponding constant speed control system is determined, including: The system monitors the constant speed status of the train in real time and determines the corresponding constant speed mapping relationship based on the identification of the constant speed status. At the same time, it collects the optimal combination of PID parameters and determines the corresponding constant speed parameters based on the optimal combination of PID parameters, the constant speed mapping relationship corresponding to the constant speed status, and the corresponding constant speed event.

9. The constant speed control method for trains based on particle swarm optimization algorithm according to claim 8, characterized in that, The constant speed state of the marked train is determined by identifying the corresponding constant speed parameters based on the optimal PID parameter combination and the train's constant speed state. A corresponding constant speed control system is then determined based on these constant speed parameters, the train's actual speed, and the corresponding operating state. The system also includes: The actual speed of the train is collected, and the speed change during the train's operation is marked. The first level of constant speed control is determined based on the actual speed of the train, the corresponding speed change, and the constant speed parameter. The train's operating status is collected, and the second level of constant speed control content is determined based on the train's actual speed, the corresponding speed change, and the train's operating status. The corresponding constant speed control system is then determined based on the first and second levels of constant speed control content.

10. A constant speed control system for trains based on particle swarm optimization algorithm, characterized in that, The constant speed control system for trains based on particle swarm optimization algorithm is applied to the constant speed control method for trains based on particle swarm optimization algorithm as described in any one of claims 1-9.

Images