Machine learning-based urban rail train energy consumption dynamic optimization method and system

Abstract

This invention discloses a machine learning-based method and system for dynamic energy consumption optimization of urban rail transit trains, relating to the field of data processing technology. The method includes: interactively acquiring train operation status data and station passenger flow data, analyzing passenger flow disturbance trends, and generating dynamic load evolution characteristics. It also involves collecting track environment data such as gradient curvature before and after stations and environmental disturbance data, combining this with load characteristics to conduct resistance correlation analysis, and obtaining dynamic resistance evolution characteristics. Furthermore, it involves constructing a dynamic energy consumption constraint field based on operational tasks, conducting dual-path optimization for pre-station kinetic energy recovery and post-station traction start control, and realizing dynamic optimization management of urban rail transit train energy consumption based on the optimization results. This invention solves the technical problem of existing urban rail transit train energy consumption optimization methods using static parameter settings and independent control processes, which cannot couple the dynamic changes of passenger flow load and track environment. It achieves the technical effect of improving energy consumption control accuracy and energy recovery utilization rate, reducing energy consumption while meeting operational requirements.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for dynamic optimization of energy consumption of urban rail trains based on machine learning. Background Technology

[0002] Energy consumption optimization for urban rail trains directly impacts operational efficiency and energy conservation and carbon reduction, making dynamic and precise control through data processing crucial. Existing technologies largely employ static parameters, fixed resistance, and load calculation models, relying on traditional data processing and experience-based settings, failing to effectively integrate multi-source data such as train operation, station passenger flow, and track environment. These methods struggle to adapt to dynamic disturbances like passenger flow fluctuations and gradient curvature, resulting in static data processing, fragmented control mechanisms, and a lack of coordination between kinetic energy recovery and traction control. Traditional technologies cannot construct precise load-resistance relationships through dynamic data processing, leading to energy consumption constraints and control optimization detached from actual operating conditions. This results in incomplete and inaccurate optimization data, failing to meet the demands of dynamic, precise, and efficient energy consumption management for urban rail trains. Summary of the Invention

[0003] This application provides a machine learning-based method and system for dynamic optimization of energy consumption of urban rail trains, which solves the technical problem that existing urban rail train energy consumption optimization methods use static parameter settings, have independent control processes, and cannot couple the dynamic changes of passenger flow load and line environment.

[0004] The first aspect of this application provides a machine learning-based method for dynamic optimization of energy consumption of urban rail transit trains. The method includes: interactively obtaining train operation status data during the current operation of the urban rail transit train; connecting to a data exchanger at a stop to obtain station passenger flow data; after reading the current load characteristics from the train operation status data, performing passenger flow disturbance trend analysis using the station passenger flow data to establish dynamic load evolution characteristics characterizing the load change trend of the urban rail transit train; obtaining track environment data based on the track before and after the stop, including gradient data, curvature data, and section environmental disturbance data; performing section running resistance correlation analysis based on the track environment data and dynamic load evolution characteristics to establish dynamic resistance evolution characteristics of the operating section; constructing a dynamic energy consumption constraint field using the target operating task and dynamic resistance evolution characteristics; performing dual-path control optimization using the dynamic energy consumption constraint field and train operation status data, including kinetic energy recovery optimization before the stop and traction start control optimization after the stop; and performing dynamic optimization management of urban rail transit train energy consumption based on the dual-path control optimization results.

[0005] The second aspect of this application provides a machine learning-based dynamic energy consumption optimization system for urban rail trains. The system includes: a train operation status data acquisition module, used to interactively acquire train operation status data of the urban rail train during its current operation; a dynamic load evolution feature construction module, used to connect to a data interface at a stop station to acquire station passenger flow data, and after reading the current load features from the train operation status data, to perform passenger flow disturbance trend analysis using the station passenger flow data to establish dynamic load evolution features characterizing the load change trend of the urban rail train; and a line environment data acquisition module, which acquires line environment data based on the line before and after the stop station, the line environment data including gradient data, curvature data, etc. The system includes: an interval environmental disturbance data module; a dynamic resistance evolution characteristic construction module, used to perform interval operation resistance correlation analysis based on the line environment data and dynamic load evolution characteristics, and establish the dynamic resistance evolution characteristics of the operating section; an operation dynamic energy consumption constraint field construction module, which constructs an operation dynamic energy consumption constraint field using the target operation task and dynamic resistance evolution characteristics; a dual-path control optimization execution module, which performs dual-path control optimization using the operation dynamic energy consumption constraint field and train operation status data, including kinetic energy recovery optimization before stopping stations and traction start control optimization after stopping stations; and a train energy consumption management execution module, used to perform dynamic optimization management of urban rail train energy consumption based on the dual-path control optimization results.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] This application acquires multi-source data on urban rail train operation, station passenger flow, and track environment. Through passenger flow disturbance trend and section operation resistance correlation analysis, it constructs dynamic load and resistance evolution characteristics. Based on the operation task, it builds an energy consumption constraint field and carries out dual-path control optimization, thereby dynamically optimizing train traction and kinetic energy recovery strategies. This makes urban rail train energy consumption management more precise and energy utilization efficiency higher, achieving the technical effect of improving energy consumption management accuracy and energy recovery utilization rate, and reducing energy consumption while meeting operational requirements. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the machine learning-based dynamic energy consumption optimization method for urban rail trains provided in this application embodiment.

[0010] Figure 2This is a schematic diagram of the structure of the machine learning-based dynamic energy consumption optimization system for urban rail trains provided in the embodiments of this application.

[0011] Explanation of reference numerals in the attached diagram: 1. Train operation status data acquisition module; 2. Dynamic load evolution feature construction module; 3. Track environment data acquisition module; 4. Dynamic resistance evolution feature construction module; 5. Dynamic energy consumption constraint field construction module; 6. Dual-path control optimization execution module; 7. Train energy consumption management execution module. Detailed Implementation

[0012] This application provides a machine learning-based method and system for dynamic optimization of energy consumption of urban rail trains, which solves the technical problem that existing urban rail train energy consumption optimization methods use static parameter settings, have independent control processes, and cannot couple the dynamic changes of passenger flow load and line environment.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, a machine learning-based method for dynamic optimization of energy consumption in urban rail transit trains includes: Interactively obtain train operation status data of urban rail trains during the current operation process.

[0016] Specifically, the onboard control unit of the urban rail train establishes real-time communication connections with various subsystems via the vehicle bus. The onboard control unit assigns communication addresses to each subsystem according to the vehicle bus standard protocol, enabling targeted data interaction with each subsystem. The vehicle bus adopts a multi-functional vehicle bus or an Ethernet train backbone network that conforms to rail transit industry standards. The multi-functional vehicle bus complies with the IEC 61375-1 standard, and the Ethernet train backbone network complies with the IEC 61375-2-5 standard. As the core node for data interaction, the onboard control unit continuously sends data request commands to each subsystem. The frame format of the data request commands strictly follows the protocol specifications of the corresponding vehicle bus.

[0017] The train traction system collects data on the output power, output torque, operating current, and voltage of the traction motor. The train braking system collects data on brake cylinder pressure, braking deceleration, and braking status. Train speed sensors are installed at the wheel axles of the train bogies to collect real-time operating speed and mileage data. The train positioning system is installed at the front and rear of the train to collect real-time position and direction of travel data. Train load sensors are installed below the air springs in the train cars to collect real-time load data, which represents the train's current total mass. The train air conditioning system collects data on air conditioning operating mode, compressor power, and fan speed. The train auxiliary power supply system collects data on auxiliary power supply output voltage, output current, and load power. For example, the sampling frequency of the speed sensors and positioning system is 10 Hz, the sampling frequency of the traction system, braking system, and auxiliary power supply system is 5 Hz, and the sampling frequency of the load sensors and air conditioning system is 1 Hz.

[0018] Each subsystem adds a CRC32 checksum to the collected real-time data and encapsulates it according to a preset communication protocol. The encapsulated data is then transmitted to the vehicle control unit via the vehicle bus. When the vehicle control unit receives the data, it first verifies the checksum. If the verification fails, it sends a data retransmission request to the corresponding subsystem. If three consecutive retransmissions fail, the data is marked as abnormal. The system then retrieves valid data from the previous and even earlier times for that data type. Based on the values ​​and time intervals of adjacent valid data, it uses linear interpolation to fill in the missing data at the current time, ensuring the continuity of the operating status data.

[0019] The onboard control unit performs timestamp synchronization processing on the received multi-source valid data. The timestamp synchronization adopts the standard time provided by the train's unified clock system. The train's unified clock system is synchronized by the Beidou satellite navigation system. The timestamp synchronization accuracy is not less than ±1 millisecond, and the synchronization period is 100 milliseconds.

[0020] The onboard control unit stores the synchronized train operation status data into local non-volatile memory according to the corresponding sampling frequency. This non-volatile memory uses flash memory or electrically erasable programmable read-only memory. The energy consumption optimization processing unit is deployed onboard in the train driver's cab control cabinet. The energy consumption optimization processing unit and the onboard control unit establish a communication connection via an Ethernet interface, using the TCP / IP protocol. The onboard control unit outputs the stored train operation status data to the energy consumption optimization processing unit in real time.

[0021] The data exchanger connected to the station obtains station passenger flow data. After reading the current load characteristics in the train operation status data, the station passenger flow data is used to perform passenger flow disturbance trend analysis and establish dynamic load evolution characteristics that characterize the load change trend of urban rail trains.

[0022] In this embodiment, the data interactor is a standard data interface device deployed at the center of the rail transit network to connect the train and platform data acquisition equipment and complete data transfer and interaction.

[0023] Optionally, firstly, station passenger flow data is collected by the platform video acquisition device. The data intercom of the rail transit network center communicates digitally with the video acquisition device and interacts with the urban rail train to provide the train with station passenger flow data. The specific implementation process of this step will be described in detail below.

[0024] Next, the energy consumption optimization processing unit obtains synchronized train operation status data from the onboard control unit, extracts real-time load data as the current load characteristic, and the real-time load data is the train's current total mass, collected by load sensors deployed under the air springs in the carriages and processed with timestamp synchronization. Passenger flow disturbance trend analysis is executed immediately after the train obtains station passenger flow data.

[0025] Next, passenger flow disturbance trend analysis is performed using station passenger flow data. By analyzing the increasing passenger flow data, the growth characteristics of the number of people waiting on the platform are established. The number of people waiting at the station is predicted by combining the train arrival time. The characteristics of boarding and alighting behavior are established by reading the historical boarding and alighting data of the stops. The net load change analysis is performed by combining the predicted number of people waiting, the characteristics of boarding and alighting behavior and the current load characteristics. Dynamic load evolution characteristics that characterize the load change trend of urban rail trains are established and dynamic load evolution characteristics within the operating section are generated. The specific implementation process of this step is also described in detail below.

[0026] The route environmental data is obtained based on the route before and after the stops. The route environmental data includes gradient data, curvature data, and section environmental disturbance data.

[0027] In one embodiment of this application, when the train is 3 kilometers before the target stop, the energy consumption optimization processing unit initiates the line environment data acquisition process. A BeiDou positioning receiver module is deployed on an unobstructed location on the train roof, with a planar positioning accuracy of no less than ±0.1 meters, an elevation measurement accuracy of no less than ±0.2 meters, and a sampling frequency of 10 Hz. An inertial measurement unit is deployed on a rigid platform in the middle of the train body, with a sampling frequency of 100 Hz and a pitch angle measurement accuracy of no less than ±0.01 degrees. The BeiDou positioning receiver module and the inertial measurement unit achieve microsecond-level time synchronization through an onboard time synchronization system. The energy consumption optimization processing unit receives the above data, uses a Kalman filter algorithm to denoise the raw data, takes two sampling points with a 0.1-second interval, calculates the ratio of the elevation difference to the horizontal distance between the two points, and obtains the line slope data for the corresponding location. The data covers a 2-kilometer line range before and after the stop. If the BeiDou positioning signal is lost, the inertial measurement unit calculates the train's position and elevation data, and simultaneously calls the slope data from the local line database for correction.

[0028] Next, the train's horizontal position coordinate data is received from the BeiDou positioning receiver module, with a sampling frequency of 10 Hz. The horizontal coordinates of three consecutive sampling points are then taken. , , The radius R of the circumcircle formed by the three points is calculated using the formula R = abc / (4S), where a, b, and c are the lengths of the three sides of the triangle, S is the area of ​​the triangle, and the curvature of the line is 1 / R. The sampling location for curvature data is synchronized with the slope data, covering a 2-kilometer range before and after each stop. All calculated slope and curvature data are compared with pre-stored data in the local line database; if the deviation exceeds 5%, the database data prevails.

[0029] The train stores a local database of the railway line environment. This database pre-loads information on tunnel lengths, curve types, and elevated sections for each section of the line, and is updated monthly from the rail transit network center via a data exchange. Wind speed and direction sensors are installed on the windward side of the train's roof, with a measurement range of 0 to 60 meters per second, a measurement accuracy of no less than ±0.5 meters per second, a response time of no more than 1 second, and a sampling frequency of 1 Hz. The energy consumption optimization processing unit retrieves track type data for the sections before and after the train's current stop based on the train's current location. For example, environmental resistance coefficients of 0.8, 1.2, and 1.0 are set for tunnels, elevated sections, and ground-level sections, respectively. These environmental resistance coefficients are multiplied by the square of the real-time wind speed to generate environmental disturbance data for the section.

[0030] The track environment data is updated every second. After each update, a CRC32 check is performed. If the check fails, the data set is discarded and the previous valid data set is used. If no valid sensor data is obtained for 10 consecutive seconds, the corresponding section's environmental data is directly retrieved from the local track database and marked as pre-stored replacement data. The complete track environment data is transmitted to the energy consumption optimization processing unit to provide basic parameters for subsequent train running resistance calculations.

[0031] Based on the line environment data and dynamic load evolution characteristics, a section operation resistance correlation analysis is performed to establish the dynamic resistance evolution characteristics of the operating section.

[0032] Specifically, based on the line environment data and dynamic load evolution characteristics, a correlation analysis of the running resistance of the section is carried out. First, various types of line environment data are converted into unit mass resistance coefficients. Combined with the dynamic load evolution characteristics, real-time mass data of the train's running position are obtained. Then, the two types of data are input into the position resistance fitting model to generate the position dynamic resistance curve, thereby establishing the dynamic resistance evolution characteristics corresponding to the running section. The specific implementation process of this step will be described in detail below.

[0033] A dynamic energy consumption constraint field is constructed by utilizing the target operation task and the dynamic resistance evolution characteristics.

[0034] Specifically, the boundary conditions are the interval running time, start and end speed constraints, and line speed limit included in the target operation task. The dynamic resistance evolution characteristics are regarded as state transition parameters. Based on dynamic programming, the feasible speed range and upper and lower energy consumption boundaries of each position point of the train are calculated, and then the dynamic energy consumption constraint field of operation is constructed. The specific implementation process of this step will be described in detail below.

[0035] The dual-path control optimization is performed using the dynamic energy consumption constraint field and train operation status data. The dual-path control optimization includes kinetic energy recovery optimization before stopping at the station and traction start control optimization after stopping at the station.

[0036] Specifically, firstly, dual-path control optimization is carried out based on the dynamic energy consumption constraint field and train operation status data. Various high-energy-consumption operation events in the section behind the station are identified and instantaneous energy demand curves are constructed. Combined with train operation-related data, a recoverable kinetic energy sequence corresponding to the station-front operation trajectory is established. The kinetic energy recovery optimization is completed through time-series matching analysis results.

[0037] Next, the available recovered energy before the stopping station is obtained. The traction demand model is constructed by combining the resistance load characteristics behind the station with the operation task to obtain the target energy demand. Based on the two types of energy data, the energy consumption constraint relationship is established and the energy release range and rate are determined. Multiple traction control candidate sequences are obtained through reverse optimization. After operation deviation analysis and energy consumption evaluation, the lowest energy consumption traction control strategy that meets the operation requirements is output. The specific implementation process of the above dual-path control optimization steps is described in detail below.

[0038] Dynamic optimization management of urban rail train energy consumption is carried out based on the optimization results of dual-path control.

[0039] Specifically, after completing the dual-path control optimization, the energy consumption optimization processing unit obtains the kinetic energy recovery optimization results before the stopping station and the traction start control optimization results after the stopping station. The kinetic energy recovery optimization result before the stopping station is the determined optimal correction scheme, including the corrected operating speed curve, the inertia entry timing delay, and the traction holding duration adjustment. The traction start control optimization result after the stopping station is the lowest energy consumption traction control strategy output, i.e., the set with the lowest overall energy consumption among the candidate traction control sequences, including traction power output parameters, traction holding duration parameters, acceleration control parameters, and inertia entry parameters.

[0040] Closed-loop adjustment of dynamic optimization management: During train operation, the energy consumption optimization processing unit monitors the actual operating speed, energy consumption, and arrival time at a frequency of 1Hz, and compares them in real time with the feasible speed range and planned operating time in the dynamic energy consumption constraint field. If the actual energy consumption deviates from the expected value by more than 5% or the delay / early arrival exceeds 2 seconds, feedback correction is triggered: For the stage before stopping at a station, the aforementioned dynamic correction optimization method for kinetic energy recovery is invoked again, and the speed curve is finely adjusted while maintaining stopping comfort and accuracy constraints; for the stage after stopping at a station, the aforementioned inverse optimization method for traction start control is invoked again, and the traction power output and coasting engagement timing are dynamically adjusted in the remaining section. The corrected control commands are immediately issued to each subsystem of the train to ensure that the energy consumption is always in the optimal range throughout the entire journey. Meanwhile, the energy consumption optimization processing unit records the input parameters, intermediate results, and execution effects (actual recovered energy, actual traction energy consumption, on-time arrival rate, parking accuracy, etc.) of this dual-path control optimization to the local non-volatile memory, and uploads them to the rail transit network center via the vehicle-to-ground communication module for subsequent periodic updates of model parameters such as high energy consumption threshold and traction efficiency coefficient, for example, once a month.

[0041] Through the above steps, the results of kinetic energy recovery optimization and traction start control optimization are transformed into real-time train control commands. Combined with a dynamic feedback correction mechanism, the dynamic optimization management of urban rail train energy consumption is fully realized.

[0042] Furthermore, the method provided in this application embodiment includes: The station passenger flow data is read through the platform video acquisition device. The data interactive device is digitally connected to the platform video acquisition device, and the data interactive device is the data interface of the rail transit network center, used to perform interactive communication with the urban rail train and the platform video acquisition device.

[0043] In this embodiment of the application, the rail transit network center is the core platform for coordinating the operation and management of the entire network and the aggregation and processing of various types of operational data.

[0044] Specifically, the platform video capture devices are deployed above the waiting areas of each stop. They use high-definition network camera equipment that meets the standards of the rail transit industry, with a resolution of no less than 1920×1080 pixels, a frame rate of 25 frames per second, a field of view of no less than 120 degrees, and an installation height of 3 to 4 meters. The video capture range covers the entire waiting area of ​​the corresponding train stop. The platform video acquisition device has a built-in image processing unit that analyzes the acquired video images frame by frame. It uses the background subtraction method, commonly used in rail transit passenger flow statistics, to extract moving targets. First, it acquires the initial 10 frames of images without moving targets, calculates the average pixel value to establish an initial background model, and calculates the difference between the current frame image and the corresponding pixel value of the background model. Pixels with a difference greater than a preset threshold are marked as foreground pixels, and the rest are background pixels. At the same time, the background model is updated in real time using the current frame image and the background model with a weight coefficient of 0.02. Then, it performs connected component analysis on the binary image composed of foreground pixels. First, it performs one erosion and one dilation operation using a 3×3 structuring element to remove isolated noise points. Then, it uses the four-neighbor labeling method to mark all connected regions, counts the pixel area of ​​each connected region, and filters out noise regions with an area smaller than a preset threshold. Each connected region that meets the area requirement corresponds to one passenger. In this way, it counts the number of people waiting in the waiting area, the increase in passenger flow per unit time, and the passenger flow distribution information, generating standardized station passenger flow data containing timestamps, station numbers, number of people waiting, and passenger flow increments.

[0045] The data exchange unit is deployed at the center of the rail transit network, serving as the unified data exchange interface for the network. It establishes an industrial Ethernet digital communication connection with the platform video acquisition devices at each station, adhering to the IEC 61375-3-1 standard. The data exchange unit receives real-time passenger flow data uploaded by the platform video acquisition devices, using a pre-configured buffer capacity to cache the data. The data retention time meets the real-time data exchange requirements of the network. Simultaneously, it performs format normalization processing on the received data, converting heterogeneous data collected from different stations into structured data containing fixed fields, uniformly using UTF-8 encoding. The station gate system and ticketing system connect to the data exchange unit via the standard communication interface of the rail transit network center, enabling them to upload real-time passenger flow data for their respective stations to the data exchange unit.

[0046] When the urban rail train reaches a distance of 1000 meters from the target stop, it establishes a wireless communication connection with the data exchanger at the rail transit network center via its onboard vehicle-to-ground communication module. The vehicle-to-ground communication uses the LTE-M rail transit-specific communication standard. The urban rail train sends a passenger flow data retrieval command to the data exchanger, containing the target station number, train number, and current time. Upon receiving the command, the data exchanger retrieves the real-time passenger flow data for the corresponding stop over the past 5 minutes, adds a CRC32 checksum to the data, and then transmits it to the urban rail train.

[0047] After receiving real-time passenger flow data, the urban rail train first verifies it. If the verification passes, the train confirms the acquisition of valid passenger flow data. If the verification fails, the train sends a data retransmission command to the data exchanger. The retransmission timeout is set to 500 milliseconds, and the maximum number of retransmissions is 3. If three consecutive retransmissions fail, a backup data acquisition process is triggered, requesting real-time passenger flow data from the corresponding station's gate system and ticketing system as supplementary data.

[0048] Finally, the urban rail train performs clock synchronization processing on the acquired valid passenger flow data. The synchronization clock uses the train's BeiDou unified time synchronization system, with a synchronization accuracy of no less than ±10 milliseconds. The urban rail train's local storage unit pre-stores historical average passenger flow data for each station's corresponding time period. Specifically, it selects data from several consecutive months of normal operating periods, categorizes time periods by weekdays, weekends, and holidays, calculates the daily passenger flow values ​​for the same station's corresponding time period, and then summarizes and archives the arithmetic average. If the train-to-ground communication is interrupted and real-time passenger flow data cannot be obtained, the urban rail train directly retrieves the corresponding station's historical average passenger flow data from its local storage and marks this data as historical substitute data. After synchronization, the station passenger flow data is transmitted to the energy consumption optimization processing unit to provide data support for subsequent passenger flow disturbance analysis.

[0049] Furthermore, the method provided in this application embodiment includes: The station passenger flow data is analyzed to obtain the increasing passenger flow data entering the target train waiting area within a unit of time, and the growth characteristics of the number of people waiting on the platform are established. The arrival time of the urban rail train is obtained, and the predicted number of people waiting for the urban rail train is predicted using the arrival time and the growth characteristics of the number of people waiting on the platform. The historical boarding and alighting data of the stops are read to establish the boarding and alighting behavior characteristics for the corresponding time period. The predicted number of people waiting, boarding and alighting behavior characteristics and current load characteristics are used to perform net load change analysis to generate dynamic load evolution characteristics within the operating section.

[0050] Optionally, the energy consumption optimization processing unit receives station passenger flow data, segments the data into fixed 10-second time intervals, calculates the difference in the number of waiting passengers within each time interval, and obtains the passenger flow increase data entering the target train waiting area per unit time. Then, the least squares method is used to linearly fit the passenger flow increase data for five consecutive time intervals, establishing a linear growth function of the number of waiting passengers on the platform over time, which serves as the characteristic of the platform waiting passenger growth. If the passenger flow growth exhibits non-linear characteristics, only the data from the most recent three time intervals are used for fitting to ensure prediction accuracy.

[0051] Then, the planned arrival time and standard operating curve data of the target section are obtained from the train operation control system. Since the train's speed varies significantly during the three stages of departure acceleration, mid-journey smooth operation, and arrival deceleration, and is not traveling at a constant speed, directly using the current speed to calculate the remaining travel time will result in a large error. Therefore, it is necessary to obtain standard operating curve data that includes the train's target speed corresponding to different positions within the section. Combined with real-time train positioning data, the train's current specific position within the section is determined, and the speed change sequence from the current position to the target station is extracted from the standard operating curve. The speed change sequence is integrated to obtain the remaining travel time from the current position to the target station, which is then used to correct the actual arrival time. The time difference between the actual arrival time and the current time is calculated, and this time difference is substituted into a linear growth function corresponding to the growth characteristics of the number of people waiting on the platform to calculate the predicted number of people waiting on the platform when the train arrives.

[0052] Next, the historical boarding and alighting data for the target stop within the same time period over the past 30 days is retrieved from the rail transit network center via a data interoperator. The same time period is defined as a period with the same date type and time interval as the current date. Date types are categorized into weekdays, weekends, and public holidays, with each time period lasting 15 minutes. Outliers in the historical data are removed using the 3σ principle. Statistics are then grouped by date type and time period to calculate the average number of passengers boarding, the average number of passengers alighting, and the ratio of boarding to alighting within that time period, establishing the boarding and alighting behavior characteristics for that time period.

[0053] Next, a preset average weight standard of 60 kg per person is adopted, which already includes the average weight of passengers' carry-on luggage. The specific value can be adjusted by those skilled in the art according to the actual situation. The predicted number of waiting passengers is converted into predicted boarding load, and the average number of disembarking passengers is converted into predicted disembarking load. The current load characteristic is subtracted from the predicted disembarking load and added to the predicted boarding load to obtain the net load after the train stops at that station. According to the station sequence of the train operation line, the net load after stopping at each station is calculated sequentially to generate the dynamic load evolution characteristics of the train throughout the entire operating section. The dynamic load evolution characteristics are generated at station intervals, updated once at each station, and fine-tuned every 30 seconds during train operation based on real-time operating data.

[0054] If the station passenger flow data is historical substitute data, then the historical average passenger flow data is directly used as the predicted waiting number. If the train is delayed or arrives early, then the predicted waiting number is recalculated and the dynamic load evolution characteristics are updated based on the difference between the actual arrival time and the planned arrival time.

[0055] The above steps enable accurate prediction of load change trends during urban rail train operation. Unlike the delayed weighing method after arrival, this method adjusts the operating parameters of the train from its current position to the station in advance based on the predicted load changes, optimizes the kinetic energy recovery strategy, avoids instantaneous energy impact after stopping and starting, and provides accurate and reliable dynamic load basis for subsequent section operation resistance calculation and energy consumption optimization control.

[0056] Furthermore, the method provided in this application embodiment includes: The gradient data, curvature data, and environmental disturbance data of the section are converted into unit mass drag coefficients respectively; real-time mass data of urban rail trains at each operating position point are established based on the dynamic load evolution characteristics; the real-time mass data and unit mass drag coefficients are used as input data and input into the position drag fitting model to establish the position dynamic drag curve, and the position dynamic drag curve is output as the dynamic drag evolution characteristics.

[0057] Specifically, the energy consumption optimization processing unit performs data conversion according to the general calculation standard for urban rail train operating resistance. Gradient data is expressed in per mille (‰), and the unit mass drag coefficient corresponding to the gradient is equal to the gradient value multiplied by 9.8 N / kg. Curvature data is expressed in curve radius in meters, and the unit mass drag coefficient corresponding to the curvature is equal to 600 divided by the curve radius. Inter-section environmental disturbance data is directly used as the corresponding unit mass drag coefficient, expressed in N / kg.

[0058] Next, the previously generated dynamic load evolution features are retrieved. These features are updated and recorded only at each stop. Continuous running position nodes are divided along the train's route at fixed intervals of 10 meters. No passenger boarding or alighting occurs between adjacent stops. All position nodes within the interval uniformly use the load values ​​corresponding to when the train departed the previous stop, serving as the real-time mass data for each running point within the interval. After the train arrives at the next stop, the updated load values ​​from that stop are used. The mass data unit is kilograms.

[0059] The location resistance fitting model was built using a multiple linear fitting method. First, at least 1000 sets of measured samples from actual urban rail train operation scenarios were collected, covering the actual operating resistance values ​​under different gradients, curvatures, environmental conditions, and passenger loads. Outliers were removed from the samples using the 3σ principle. The gradient unit mass resistance coefficient, curvature unit mass resistance coefficient, environmental unit mass resistance coefficient, and the real-time train mass were set as model input variables for each sample, and the actual operating resistance corresponding to the sample was set as the model output variable. The mathematical form of the model is: Operating Resistance = (Gradient Unit Mass Resistance Coefficient + Curvature Unit Mass Resistance Coefficient + Environmental Unit Mass Resistance Coefficient + Basic Resistance Coefficient) × Mass, where the basic resistance coefficient is an inherent parameter of the train, with a value of 2.5 N / kg. The least squares method was used to iteratively calculate the coefficients in the model. The iteration terminated when the sum of squared residuals was less than 1e-6. After the parameters were calculated, the model structure was finalized and stored in the energy consumption optimization processing unit, updated quarterly using the latest operating data from the rail transit network center.

[0060] Finally, the real-time mass data of all points along the entire line and the converted three types of unit mass resistance coefficients are input into the finalized position resistance fitting model. The model calculates the corresponding operating resistance at each point according to the predetermined calculation logic, with the unit of operating resistance being Newtons (N). The resistance calculation results of all points are integrated according to the order of train operation, forming a discrete point sequence with 10-meter intervals. These discrete points are then arranged to form a complete position dynamic resistance curve, which serves as the dynamic resistance evolution characteristic for the current operating section.

[0061] If environmental data for a certain type of line is missing or abnormal, the historical average resistance coefficient of the corresponding section in the local line database will be used as a substitute. If the deviation between the model calculation result and the empirical formula calculation result exceeds 10%, the empirical formula calculation result will be used and marked as abnormal data.

[0062] By combining a clear rule for converting resistance coefficients with a standardized multiple linear fitting model, the dynamic resistance values ​​at each point along the entire train journey are accurately calculated, fully presenting the resistance variation pattern within the operating section. This provides reliable basic parameters for subsequent accurate calculation and dynamic optimization control of train energy consumption.

[0063] Furthermore, the method provided in this application embodiment includes: Using the section running time, start and end speed constraints, and line speed limit in the target operation task as boundary conditions, and the dynamic resistance evolution characteristics as state transition parameters, the feasible speed range and corresponding upper and lower energy consumption boundaries of the urban rail train at each location point are solved based on dynamic programming to establish an operational dynamic energy consumption constraint field.

[0064] Specifically, the first step is to extract the limiting parameters within the target operation task, defining the allowed running time for the entire route in seconds, the initial and final speeds at the train's departure and arrival points in meters per second, and the maximum speed limits set at various points along the entire line in meters per second. The line speed limit is a variable that varies with location, with each location node corresponding to a maximum allowed speed value. All parameters are then fixed as boundary constraints for subsequent calculations.

[0065] The generated dynamic resistance evolution characteristics are retrieved, and the corresponding resistance values ​​are extracted sequentially according to the running position nodes of the line, with the running position nodes divided at fixed intervals of 10 meters. The resistance values ​​corresponding to different points are set as state transition reference parameters during train movement to characterize the force changes during train position changes.

[0066] Next, a dynamic programming recursive solution method is used for computation. The state variables are defined as the train's current node number and current speed. The decision variable is defined as the acceleration value chosen by the train at the current node, with an example acceleration value ranging from -1.5. To +1.0 Where negative values ​​represent braking and positive values ​​represent traction. The optimization objective is to calculate the minimum and maximum energy consumption for each state, while satisfying all boundary conditions.

[0067] The state transition equation is: Next node velocity = Current node velocity + Acceleration × Time interval, where Time interval = 10 ÷ (Current node velocity + Next node velocity) ÷ 2. Energy consumption is calculated using a piecewise integration method: Traction energy consumption = Traction force × Displacement distance ÷ Traction system efficiency; Braking energy consumption = Braking force × Displacement distance × Braking energy recovery rate. The traction system efficiency is taken from the fixed performance parameters calibrated at the train's factory, and the braking energy recovery rate is taken from the rated operating parameters of the onboard energy storage equipment, with values ​​set to 0.85 and 0.7 respectively for conventional applications. Traction energy consumption is calculated when acceleration is positive, and braking energy recovery consumption is calculated when acceleration is negative.

[0068] Then, a reverse recursive approach is used for calculation, starting from the train arrival position node and using the arrival termination speed as the initial state, progressively calculating backward to the departure position node. At each node, all possible acceleration values ​​are traversed, and the corresponding next node state and energy consumption value are calculated. Simultaneously, it is checked whether the line speed limit constraint and total running time constraint are met. The total running time constraint is achieved by accumulating the time intervals between each node and comparing them with the allowed running time. All states that meet the constraints are retained, and the minimum and maximum energy consumption corresponding to each state are recorded to determine the feasible speed range and upper and lower energy consumption boundaries for each position node.

[0069] By integrating the speed ranges and energy consumption boundary values ​​of all operational nodes according to the route's progression sequence, and correlating and matching spatial location, feasible speed range, and energy consumption fluctuation range, a three-dimensional data structure of dynamic energy consumption constraint field is formed. In this constraint field, each location node corresponds to a speed range, and each speed value corresponds to an energy consumption range.

[0070] By clarifying the complete structure and energy consumption calculation method of the dynamic programming model, and combining the line operation constraints to calculate the speed and energy consumption range point by point, a precise and complete dynamic energy consumption constraint field is established, which defines a scientific and reasonable operating range for the selection of train speed strategy and the optimal control of energy consumption.

[0071] Furthermore, the method provided in this application embodiment includes: Based on the dynamic energy consumption constraint field, high-energy-consuming operation events in the operating section after the stop are identified. These high-energy-consuming operation events include uphill operation events, time-limited speed-up operation events, and high traction load operation events. Instantaneous energy demand curves for the target operating section after the stop are established using these high-energy-consuming operation events. The correlation between the operating trajectory before the stop and recoverable kinetic energy is established based on the current operating speed, current load status, and remaining operating distance at the stop from the train's operating status data, thus establishing recoverable kinetic energy sequences corresponding to different operating trajectories. A time-series matching analysis is performed based on the recoverable kinetic energy sequences and the instantaneous energy demand curves, and kinetic energy recovery optimization is executed based on the time-series matching analysis results.

[0072] Specifically, the energy consumption optimization processing unit retrieves the generated dynamic energy consumption constraint field and extracts the upper and lower boundary values ​​of energy consumption for each node according to its operating position on the line. The high energy consumption threshold is determined by statistically analyzing historical operating data of the same type of train on the same line over the past three months. The average energy consumption value of all 10-meter operating position nodes is calculated, and 1.5 times this average value is taken as the high energy consumption threshold. The upper limit of energy consumption for each node is compared with the high energy consumption threshold. When the upper limit of energy consumption for a certain operating position node exceeds the high energy consumption threshold, the event is classified based on the corresponding line environment data. For example, if the gradient unit mass drag coefficient corresponding to the node is greater than 0.02 N / kg, it is determined as an uphill operation event. If the target operating speed corresponding to the node increases by more than 2 meters per second compared to the previous node and the remaining operating time is less than 80% of the planned remaining time, it is determined as a timed speed-up operation event. If the real-time mass corresponding to the node is greater than 1.8 times the empty mass of the train and the acceleration is greater than 0.5... If this occurs, it is determined to be a high traction load operation event. The train's unloaded mass is a fixed parameter calibrated at the factory and is pre-stored in the energy consumption optimization processing unit.

[0073] Next, all identified high-energy-consuming operation events are arranged according to their position on the line, and the starting and ending nodes for each event are determined. When the energy consumption upper limit of multiple consecutive nodes exceeds the high-energy-consuming threshold, the first node to exceed the threshold is the starting node of the event, and the last node to exceed the threshold is the ending node. If the number of nodes between two adjacent high-energy-consuming events does not exceed three, the two events are merged into a single continuous high-energy-consuming event. This interval is a conventional design value determined based on line operation experience. Based on the energy consumption values ​​of the corresponding nodes in the dynamic energy consumption constraint field, the instantaneous energy demand of each high-energy-consuming operation event within its duration is calculated. Instantaneous energy demand = energy consumption upper limit of the node × node interval distance ÷ train target operating speed at the node. The instantaneous energy demand for non-high-energy-consuming event intervals is taken as the average energy consumption value of each node within that interval. With time as the horizontal axis and instantaneous energy demand as the vertical axis, the calculation results of all nodes are sequentially connected using linear interpolation to form the instantaneous energy demand curve for the target operating section after stopping at the station.

[0074] Next, the train's current operating status data is acquired, including current speed, current load status, and remaining distance to the stop. Starting from the train's current position and ending at the stop, multiple different deceleration trajectories are generated. Each trajectory corresponds to a different braking deceleration value; an example braking deceleration value range is -0.5. To -1.5 This range is taken from the braking performance parameter range specified by the train manufacturer, with intervals of 0.1. Eleven different deceleration trajectories were generated. Based on the speed change process of each trajectory, the kinetic energy of the train at each time point was calculated at 0.1-second intervals: kinetic energy = 0.5 × real-time mass × speed squared. Combined with the braking energy recovery rate, the recoverable kinetic energy value at each time point during braking for each trajectory was calculated and arranged in chronological order to form a sequence of recoverable kinetic energy corresponding to different trajectories.

[0075] Finally, a unified timeline is established to complete data alignment. Based on the target operating speed corresponding to each location node in the dynamic energy consumption constraint field, the transit time of each node is calculated, and the timestamp corresponding to each node is accumulated, converting the location-based instantaneous energy demand curve into a time-based curve. All recyclable kinetic energy sequences are aligned with the converted instantaneous energy demand curves on the unified timeline. The difference between recyclable kinetic energy and instantaneous energy demand at each time point is calculated, and the absolute value of the difference at all time points is accumulated to obtain the matching error of each operating trajectory. The matching error threshold is set to 10% of the total energy demand of the target operating segment. The operating trajectory with the smallest matching error and less than the matching error threshold is selected as the optimal matching scheme, and the correspondence between the recyclable kinetic energy sequence and the instantaneous energy demand curve corresponding to this scheme is recorded as the result of the time-series matching analysis. If the matching error of all operating trajectories exceeds the preset threshold, the operating trajectory with the largest total recyclable kinetic energy is selected as the alternative scheme.

[0076] By employing clear parameter statistical methods and event judgment rules, and combining energy data alignment and error minimization matching along a unified time axis, precise timing matching between train braking energy recovery and subsequent high-energy-consumption events in the section is achieved, providing a reliable core matching basis for subsequent dual-objective control optimization.

[0077] Furthermore, the method provided in this application embodiment includes: Determine whether there is an energy gap interval in the time-series matching analysis results; if there is an energy gap interval in the time-series matching analysis results, establish a target kinetic energy recovery threshold that satisfies the energy gap interval; under the condition of satisfying the preset docking comfort constraints and docking accuracy constraints, use the target kinetic energy recovery threshold to perform dynamic correction and optimization of the running trajectory before the docking station, the dynamic correction and optimization including correcting the running speed, delaying the inertial entry timing, and adjusting the traction holding time; complete the kinetic energy recovery optimization based on the dynamic correction and optimization results.

[0078] In one embodiment, the energy consumption optimization processing unit retrieves the time-series matching analysis results and compares the recoverable kinetic energy sequence with the instantaneous energy demand curve point by point on a unified time axis. When the instantaneous energy demand at a certain time point is greater than the corresponding recoverable kinetic energy value, that time point is marked as an energy gap point. Consecutive energy gap points are integrated into energy gap intervals, and the sum of the differences between instantaneous energy demand and recoverable kinetic energy within each energy gap interval is calculated to obtain the total energy gap for that interval. The total energy gap of all energy gap intervals is accumulated to obtain the overall energy gap value for the target operating segment.

[0079] If the overall energy deficit is greater than zero, an energy deficit range is identified. A target recoverable kinetic energy threshold is established based on the overall energy deficit, which is equal to the overall energy deficit multiplied by 1.1, with a 10% energy margin reserved to handle fluctuations during actual operation. This threshold represents the additional amount of braking and recoverable kinetic energy required to cover the energy demands of subsequent high-energy-consumption operational events.

[0080] Under the condition of meeting preset stopping comfort and stopping accuracy constraints, dynamic correction and optimization of the train's trajectory before stopping at the station are carried out. For example, the stopping comfort constraint is that the train's longitudinal acceleration change rate does not exceed 0.5. The stopping accuracy constraint is that the error between the train's stopping position and the target position does not exceed 0.3 meters. Dynamic correction optimization is achieved through three methods working together, with the priority of the three methods being, in order: correcting the running speed, delaying the inertial engagement timing, and adjusting the traction holding duration. When used in combination, these methods are superimposed according to their priority. The specific implementation method is as follows: The first method involves adjusting the operating speed. The train's maximum braking distance is calculated based on the train's factory braking parameters and the line's maximum operating speed. The initial speed increase is set at 1.2 times the train's maximum braking distance, corresponding to a range of 100 to 300 meters before the station. Provided the total travel time within the section does not exceed the permissible travel time, the target operating speed within this range is appropriately increased. The speed increase is capped at 10% of the line speed limit and does not exceed 2 meters per second, with a step size of 0.2 meters per second. This operation utilizes the train's kinetic energy as a temporary energy carrier. By increasing the initial kinetic energy before braking, the total recovered energy is increased. This does not create additional energy consumption but rather stores the surplus energy before the station in the onboard energy storage system for use during high-energy-consumption events after the station. The prerequisite for implementation is that the train is equipped with a high-efficiency onboard energy storage system capable of maximizing the recovery and stable release of braking kinetic energy.

[0081] The second method involves delaying the inertia entry timing. The original setting for the inertia entry timing is 500 meters from the station, with a delay range of 0.5 to 2 seconds and a step size of 0.5 seconds. This range is determined based on the train's traction characteristics and passenger comfort requirements. By delaying the originally set inertia start point by a corresponding time, the duration of the traction phase is extended, allowing the train to reach a higher speed when entering inertia operation, thereby increasing the recoverable kinetic energy during braking.

[0082] The third method involves adjusting the traction holding time. The initial setting for the traction holding time is 2 seconds, with an adjustment range of 0 to 3 seconds in increments of 0.5 seconds. Appropriately extending the holding time during the traction phase, without exceeding 0.5-0.5 seconds of the maximum permissible acceleration, is also beneficial. Under the premise of allowing the train to reach a higher operating speed before switching to braking, the maximum permissible acceleration is the maximum comfort traction acceleration specified by the train manufacturer.

[0083] Next, different combinations of corrected parameters are traversed in priority order: first, the speed increase, then the inertial entry delay time, and finally the traction hold duration. A corresponding corrected trajectory is generated for each parameter combination. The speed of each node on the corrected trajectory is compared with the feasible speed range in the dynamic energy consumption constraint field to ensure that all node speeds are within the feasible range. The basic energy consumption of the original trajectory is directly extracted from the energy consumption data stored at the corresponding node in the dynamic energy consumption constraint field. The total energy consumption equals the basic energy consumption of the original trajectory plus the additional traction energy consumption brought about by the speed increase, minus the increased braking and recovery energy consumption.

[0084] Simultaneously, it was verified whether the trajectory met the stopping comfort constraints, stopping accuracy constraints, and total running time constraints for the interval. The stopping comfort constraint was verified by calculating the rate of change of longitudinal acceleration at each time point of the corrected trajectory, ensuring that the value at all time points did not exceed 0.5. The parking accuracy constraint is verified by calculating the parking position based on the corrected braking deceleration and initial velocity, ensuring that the error between the parking position and the target position does not exceed 0.3 meters. All running trajectories that meet the constraints and whose total recoverable kinetic energy is greater than or equal to the target recoverable kinetic energy threshold are selected, and the trajectory with the lowest total energy consumption is chosen as the optimal correction scheme.

[0085] If all corrective parameter combinations fail to meet the target recovered kinetic energy threshold, the target recovered kinetic energy threshold will be reduced to 90% of the original threshold. The overhead contact system auxiliary power supply will be activated according to the conventional power supply control logic for rail transit to make up for the remaining energy shortfall. If the speed increase results in insufficient total running time in the section, the subsequent inertial running time will be shortened, and the braking deceleration will be appropriately increased, but the braking deceleration must not exceed -1.5. Comfort constraints. If the remaining capacity of the on-board energy storage system is insufficient, the additional kinetic energy recovery is limited to no more than 80% of the remaining capacity of the energy storage system, and the excess is directly fed back to the grid.

[0086] Finally, the operating control parameters before the train stops at the station are adjusted according to the optimal correction scheme, and the traction and braking commands are updated in real time to complete the kinetic energy recovery optimization process.

[0087] By accurately identifying energy gaps and dynamically correcting multi-dimensional operating trajectories, combined with the energy spatiotemporal transfer mechanism of the onboard energy storage system, the system can accurately match the energy recovered during braking with the energy demand of subsequent high-energy-consumption events, significantly improving the overall energy-saving effect of the train while ensuring on-time performance and passenger comfort.

[0088] Furthermore, the method provided in this application embodiment includes: The system acquires the available recovered energy before the stopping point; it establishes a segment traction demand model for the operating section after the stopping point using the dynamic resistance evolution characteristics, dynamic load evolution characteristics, and target operating tasks, and outputs the target energy output demand that meets the target operating tasks; it establishes the energy consumption constraint relationship of the operating section after the stopping point based on the available recovered energy and target energy output demand, and determines the target release interval and target release rate of the available recovered energy in the target operating section; it performs inverse optimization analysis of traction control parameters based on the target release interval, target release rate, and dynamic resistance evolution characteristics to obtain candidate traction control sequences that meet the target operating task conditions, including traction power output parameters, traction holding duration parameters, acceleration control parameters, and coasting entry parameters; it performs segment operation deviation analysis and comprehensive energy consumption evaluation on the candidate traction control sequences, and outputs the lowest energy consumption traction control strategy that meets the target operating task conditions.

[0089] Optionally, the energy consumption optimization processing unit reads the energy data retained after braking at the stop from the on-board energy storage module, using this as the initial statistical value of recovered energy. It then deducts the energy lost due to its own circuitry according to the conventional energy loss ratio of the on-board energy storage device. This loss ratio is fixed at 3%, a commonly used empirical value for urban rail energy storage systems. After deduction, the actual usable recovered energy before the current stop is obtained, completing the energy value acquisition process.

[0090] The dynamic resistance and load evolution characteristics of the operating section following the stopping station are retrieved, and the preset target operating task content is loaded. The target operating task includes three basic indicators: the allowable running time of the section, the maximum operating speed at all points along the line, and the final stopping position. A segmented, point-by-point computational power calculation modeling method is adopted, selecting a calculation node every ten meters along the line, and extracting the corresponding operating resistance value and real-time train mass value for each node. The basic energy consumed by the train to overcome the resistance at each point is calculated sequentially. Combined with the additional energy consumption values ​​corresponding to the climbing and speed-up conditions within the section, the energy consumption results of all nodes are accumulated to obtain the target energy output requirement that the train must meet from the current station to the next station.

[0091] Next, the acquired usable recovered energy is compared with the calculated target energy output demand to establish a constraint relationship for energy consumption in the section behind the stop. Different driving sections are divided according to the route node arrangement order, with uphill sections and acceleration sections with high resistance values ​​designated as priority energy release zones. The basic energy release rate is obtained by evenly distributing the energy based on the total running time of each section. The rate is then slightly adjusted according to the resistance changes in each section, ensuring that the total cumulative energy release does not exceed the upper limit of usable recovered energy. Finally, the target release zone and matching target release rate for each route segment are determined.

[0092] Next, based on the determined target release interval and target release rate, and combined with the dynamic resistance evolution characteristics of the entire line, a reverse optimization analysis of the traction control parameters is conducted. Specifically, taking the train's stationary state at the next station as the starting point, the operating control parameters are calculated backwards towards the current departure station. The traction power output range is set with reference to the conventional operating range of urban rail train traction equipment, combined with traction holding parameters of different durations and acceleration parameters conforming to comfortable driving standards, while simultaneously changing the position node parameters corresponding to coasting initiation. Each set of parameters is combined to form a complete control logic, generating multiple candidate traction control sequences that meet the basic driving conditions. Each sequence uniformly includes traction power output parameters, traction holding duration parameters, acceleration control parameters, and coasting initiation parameters.

[0093] Finally, section operation deviation analysis and comprehensive energy consumption evaluation were conducted on all candidate traction control sequences. Each candidate traction control sequence was substituted into the line operation process to simulate the entire train's journey. Actual travel time, real-time speed, and final stopping position were recorded. The simulated data was compared with the target operation task standard data to calculate the deviation value. If the deviation was within the allowable fluctuation range, the sequence was deemed qualified. Simultaneously, the traction energy consumption and braking recovery energy generated during the operation of each sequence were statistically analyzed and calculated to obtain the section's comprehensive energy consumption value. From all qualified candidate traction control sequences, the energy consumption values ​​were compared, and the sequence with the lowest comprehensive energy consumption value was selected as the lowest energy consumption traction control strategy that meets the target operation task.

[0094] By recovering energy extraction modeling and inverse parameter optimization screening, and relying on energy consumption constraints to match driving condition parameters, a train traction control strategy that balances driving regulations and optimal energy consumption is finally obtained.

[0095] Furthermore, the method provided in this application embodiment includes: Establish a control evaluation feedback mechanism based on the dual-path control optimization results, and use the control evaluation feedback to perform feedback correction management for the current urban rail train and the current operating section.

[0096] In one embodiment, real-time data collection is performed on various operational data during the implementation of the dual-path control optimization scheme, while simultaneously retrieving theoretical operational reference data generated by the optimization calculation. The collected data includes real-time train speed, real-time traction power output, actual energy consumption per section, total travel time, and final stopping position. The collected actual operational data is then compared item by item with the preset theoretical reference data, and deviations in speed, energy consumption, duration, and stopping position are calculated. All individual deviations are then summarized to obtain the overall operational deviation of the control scheme.

[0097] The control effectiveness was assessed based on various deviation values, and the results of this dual-track control optimization were comprehensively evaluated according to three dimensions: energy consumption management level, on-time operation, and stable operation. Standardized control evaluation feedback was generated based on the evaluation results. The inherent operating performance parameters of the current train and the basic parameters of the current running section were extracted, and parameter correction calculations were performed in conjunction with the generated control evaluation feedback. For control items with deviations exceeding a reasonable range, the parameter judgment criteria and value selection ranges within the subsequent optimization process were adjusted according to the actual deviation magnitude.

[0098] Finally, the corrected reference parameters are stored in the energy consumption optimization processing unit, replacing the original calculation reference data. When the train subsequently operates in the current section or performs tasks under the same conditions, the updated reference parameters are used to perform dual-path control optimization calculations, completing the closed-loop feedback correction management corresponding to the train and the operating section.

[0099] By evaluating and providing feedback on actual operational results and iteratively correcting control parameters, we continuously optimize the adaptability of the best solution and steadily improve the quality of energy consumption optimization and management for urban rail train sections.

[0100] In summary, the machine learning-based dynamic energy consumption optimization method for urban rail trains provided in this application has the following technical effects: This application utilizes dynamic energy consumption constraints and train operation status data to perform dual-path control optimization. It identifies various high-energy-consumption operational events behind stations and constructs instantaneous energy demand curves. It establishes a recoverable kinetic energy sequence corresponding to the station-front trajectory and completes time-series matching. Combined with energy gap analysis, it optimizes kinetic energy recovery. By calculating energy supply and demand relationships, traction control parameters are derived, the optimal energy-consumption traction strategy is selected, and an evaluation feedback system is established to iteratively correct operating parameters. This effectively optimizes the energy consumption detection and control accuracy of urban rail trains, achieving the technical effect of improving energy consumption management accuracy and energy recovery utilization rate, while reducing energy consumption while meeting operational requirements.

[0101] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a machine learning-based dynamic energy consumption optimization system for urban rail trains, the system comprising: Train operation status data acquisition module 1 is used to interactively obtain train operation status data of urban rail trains during the current operation process.

[0102] The dynamic load evolution feature construction module 2 is used to connect to the data intercom of the stopping station to obtain station passenger flow data. After reading the current load features in the train operation status data, it uses the station passenger flow data to perform passenger flow disturbance trend analysis and establish dynamic load evolution features that characterize the load change trend of urban rail trains.

[0103] The line environment data acquisition module 3 acquires line environment data based on the line before and after the stop, and the line environment data includes gradient data, curvature data, and section environmental disturbance data.

[0104] The dynamic resistance evolution feature construction module 4 is used to perform section operation resistance correlation analysis based on the line environment data and dynamic load evolution features to establish the dynamic resistance evolution features of the operating section.

[0105] The dynamic energy consumption constraint field construction module 5 is used to construct the dynamic energy consumption constraint field by utilizing the target operation task and the dynamic resistance evolution characteristics.

[0106] The dual-path control optimization execution module 6 uses the dynamic energy consumption constraint field and train operation status data to perform dual-path control optimization, which includes kinetic energy recovery optimization before stopping at the station and traction start control optimization after stopping at the station.

[0107] Train energy consumption management execution module 7 is used to perform dynamic optimization management of urban rail train energy consumption based on the dual-path control optimization results.

[0108] Furthermore, the dual-path control optimization execution module 6 is used to perform the following steps: Based on the dynamic energy consumption constraint field, high-energy-consuming operation events in the operating section after the stop are identified. These high-energy-consuming operation events include uphill operation events, time-limited speed-up operation events, and high traction load operation events. Instantaneous energy demand curves for the target operating section after the stop are established using these high-energy-consuming operation events. The correlation between the operating trajectory before the stop and recoverable kinetic energy is established based on the current operating speed, current load status, and remaining operating distance at the stop from the train's operating status data, thus establishing recoverable kinetic energy sequences corresponding to different operating trajectories. A time-series matching analysis is performed based on the recoverable kinetic energy sequences and the instantaneous energy demand curves, and kinetic energy recovery optimization is executed based on the time-series matching analysis results.

[0109] Furthermore, the dual-path control optimization execution module 6 is used to perform the following steps: Determine whether there is an energy gap interval in the time-series matching analysis results; if there is an energy gap interval in the time-series matching analysis results, establish a target kinetic energy recovery threshold that satisfies the energy gap interval; under the condition of satisfying the preset docking comfort constraints and docking accuracy constraints, use the target kinetic energy recovery threshold to perform dynamic correction and optimization of the running trajectory before the docking station, the dynamic correction and optimization including correcting the running speed, delaying the inertial entry timing, and adjusting the traction holding time; complete the kinetic energy recovery optimization based on the dynamic correction and optimization results.

[0110] Furthermore, the dual-path control optimization execution module 6 is used to perform the following steps: The system acquires the available recovered energy before the stopping point; it establishes a segment traction demand model for the operating section after the stopping point using the dynamic resistance evolution characteristics, dynamic load evolution characteristics, and target operating tasks, and outputs the target energy output demand that meets the target operating tasks; it establishes the energy consumption constraint relationship of the operating section after the stopping point based on the available recovered energy and target energy output demand, and determines the target release interval and target release rate of the available recovered energy in the target operating section; it performs inverse optimization analysis of traction control parameters based on the target release interval, target release rate, and dynamic resistance evolution characteristics to obtain candidate traction control sequences that meet the target operating task conditions, including traction power output parameters, traction holding duration parameters, acceleration control parameters, and coasting entry parameters; it performs segment operation deviation analysis and comprehensive energy consumption evaluation on the candidate traction control sequences, and outputs the lowest energy consumption traction control strategy that meets the target operating task conditions.

[0111] Furthermore, the dynamic load evolution feature construction module 2 is used to perform the following steps: The station passenger flow data is analyzed to obtain the increasing passenger flow data entering the target train waiting area within a unit of time, and the growth characteristics of the number of people waiting on the platform are established. The arrival time of the urban rail train is obtained, and the predicted number of people waiting for the urban rail train is predicted using the arrival time and the growth characteristics of the number of people waiting on the platform. The historical boarding and alighting data of the stops are read to establish the boarding and alighting behavior characteristics for the corresponding time period. The predicted number of people waiting, boarding and alighting behavior characteristics and current load characteristics are used to perform net load change analysis to generate dynamic load evolution characteristics within the operating section.

[0112] Furthermore, the dynamic resistance evolution feature construction module 4 is used to perform the following steps: The gradient data, curvature data, and environmental disturbance data of the section are converted into unit mass drag coefficients respectively; real-time mass data of urban rail trains at each operating position point are established based on the dynamic load evolution characteristics; the real-time mass data and unit mass drag coefficients are used as input data and input into the position drag fitting model to establish the position dynamic drag curve, and the position dynamic drag curve is output as the dynamic drag evolution characteristics.

[0113] Furthermore, the dynamic energy consumption constraint field construction module 5 is used to perform the following steps: Using the section running time, start and end speed constraints, and line speed limit in the target operation task as boundary conditions, and the dynamic resistance evolution characteristics as state transition parameters, the feasible speed range and corresponding upper and lower energy consumption boundaries of the urban rail train at each location point are solved based on dynamic programming to establish an operational dynamic energy consumption constraint field.

[0114] Furthermore, the dynamic load evolution feature construction module 2 is used to perform the following steps: The station passenger flow data is read through the platform video acquisition device. The data interactive device is digitally connected to the platform video acquisition device, and the data interactive device is the data interface of the rail transit network center, used to perform interactive communication with the urban rail train and the platform video acquisition device.

[0115] Furthermore, the dual-path control optimization execution module 6 is used to perform the following steps: Establish a control evaluation feedback mechanism based on the dual-path control optimization results, and use the control evaluation feedback to perform feedback correction management for the current urban rail train and the current operating section.

[0116] The machine learning-based dynamic energy consumption optimization system for urban rail trains provided in this embodiment of the invention can execute the machine learning-based dynamic energy consumption optimization method for urban rail trains provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0117] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0118] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A machine learning-based method for dynamic optimization of energy consumption in urban rail transit trains, characterized in that, The method includes: Interactively obtain train operation status data of urban rail trains during the current operation process; The data exchanger connected to the stop station obtains station passenger flow data. After reading the current load characteristics in the train operation status data, the passenger flow data is used to perform passenger flow disturbance trend analysis and establish dynamic load evolution characteristics that characterize the load change trend of urban rail trains. Based on the route before and after the stops, the route environment data is obtained, including gradient data, curvature data, and section environmental disturbance data; Based on the line environment data and dynamic load evolution characteristics, perform section operation resistance correlation analysis to establish the dynamic resistance evolution characteristics of the operating section; A dynamic energy consumption constraint field is constructed by utilizing the target operation task and the dynamic resistance evolution characteristics; The dual-path control optimization is performed using the dynamic energy consumption constraint field and train operation status data. The dual-path control optimization includes kinetic energy recovery optimization before stopping at the station and traction start control optimization after stopping at the station. Dynamic optimization management of urban rail train energy consumption is carried out based on the optimization results of dual-path control.

2. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, The dual-path control optimization is performed using the aforementioned dynamic energy consumption constraint field and train operation status data, including: Based on the dynamic energy consumption constraint field, high energy consumption operation events in the operating section after the stop are identified. These high energy consumption operation events include uphill operation events, time-limited acceleration operation events, and high traction load operation events. Establish the instantaneous energy demand curve of the target operating section after stopping at a stop using high-energy-consumption operation events; Based on the current operating speed, current load status, and remaining operating distance to the stop in the train operation status data, establish the correlation between the operating trajectory before the stop and the recoverable kinetic energy, and establish the recoverable kinetic energy sequence corresponding to different operating trajectories; A time-series matching analysis is performed based on the recoverable kinetic energy sequence and the instantaneous energy demand curve, and kinetic energy recovery optimization is performed based on the time-series matching analysis results.

3. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 2, characterized in that, Based on the time-series matching analysis results, kinetic energy recovery optimization is performed, including: Determine whether the time-series matching analysis results contain an energy gap interval; If the time-series matching analysis results show an energy gap interval, then a target kinetic energy recovery threshold that satisfies the energy gap interval is established. Under the conditions of meeting the preset docking comfort constraints and docking accuracy constraints, the running trajectory before the docking station is dynamically corrected and optimized using the target recovered kinetic energy threshold. The dynamic correction and optimization includes correcting the running speed, delaying the inertial entry timing, and adjusting the traction holding time. The optimization of kinetic energy recovery is completed based on the dynamic correction optimization results.

4. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, The method of performing dual-path control optimization using the aforementioned dynamic energy consumption constraint field and train operation status data also includes: Obtain usable recovered energy before stopping at the docking station; By utilizing the dynamic resistance evolution characteristics, dynamic load evolution characteristics, and target operation tasks after stopping at the station, a section traction demand model for the operation section after stopping at the station is established, and the target energy output demand that meets the target operation task is output. Based on the available recoverable energy and target energy output requirements, establish the energy consumption constraint relationship of the operating section after the docking station, and determine the target release range and target release rate of available recoverable energy in the target operating section; Based on the target release range, target release rate and dynamic resistance evolution characteristics, reverse optimization analysis of traction control parameters is performed to obtain candidate traction control sequences that meet the target operation task conditions. The candidate traction control sequences include traction power output parameters, traction holding duration parameters, acceleration control parameters and coasting entry parameters. Perform segment operation deviation analysis and comprehensive energy consumption evaluation on the candidate traction control sequence, and output the lowest energy consumption traction control strategy that meets the target operation task conditions.

5. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, By utilizing station passenger flow data to perform passenger flow disturbance trend analysis, dynamic load evolution characteristics characterizing the load change trend of urban rail trains are established, including: Perform platform waiting passenger flow analysis on the passenger flow data of the station to obtain the passenger flow increase data entering the waiting area of ​​the target train per unit time and establish the growth characteristics of the number of people waiting on the platform; The arrival time of urban rail trains is obtained, and the predicted number of people waiting at the platform is predicted using the arrival time and the growth characteristics of the number of people waiting at the platform. Read historical boarding and alighting data from bus stops to establish boarding and alighting behavior characteristics for the corresponding time periods; By utilizing the predicted number of waiting passengers, boarding and alighting behavior characteristics, and current load characteristics, a net load change analysis is performed to generate dynamic load evolution characteristics within the operating section.

6. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, Based on the aforementioned line environment data and dynamic load evolution characteristics, a section operating resistance correlation analysis is performed to establish the dynamic resistance evolution characteristics of the operating section, including: The slope data, curvature data, and interval environmental disturbance data are converted into unit mass drag coefficients respectively. Based on the dynamic load evolution characteristics, real-time quality data of urban rail trains at each operating location point are established; The real-time mass data and the unit mass drag coefficient are used as input data and fed into the position drag fitting model to establish the position dynamic drag curve. The position dynamic drag curve is then output as the dynamic drag evolution feature.

7. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 6, characterized in that, A dynamic energy consumption constraint field is constructed using the target operational task and the dynamic resistance evolution characteristics, including: Using the section running time, start and end speed constraints, and line speed limit in the target operation task as boundary conditions, and the dynamic resistance evolution characteristics as state transition parameters, the feasible speed range and corresponding upper and lower energy consumption boundaries of the urban rail train at each location point are solved based on dynamic programming to establish an operational dynamic energy consumption constraint field.

8. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, The station passenger flow data is read through the platform video acquisition device. The data interactive device is digitally connected to the platform video acquisition device, and the data interactive device is the data interface of the rail transit network center, used to perform interactive communication with the urban rail train and the platform video acquisition device.

9. The machine learning-based dynamic energy consumption optimization method for urban rail trains as described in claim 1, characterized in that, Establish a control evaluation feedback mechanism based on the dual-path control optimization results, and use the control evaluation feedback to perform feedback correction management for the current urban rail train and the current operating section.

10. A machine learning-based dynamic energy consumption optimization system for urban rail trains, characterized in that, The system is used to implement the machine learning-based dynamic energy consumption optimization method for urban rail transit trains according to any one of claims 1-9, the system comprising: The train operation status data acquisition module is used to interactively obtain the train operation status data of the urban rail train during the current operation process; The dynamic load evolution feature construction module is used to connect to the data intercom at the stops to obtain station passenger flow data. After reading the current load features in the train operation status data, it uses the station passenger flow data to perform passenger flow disturbance trend analysis and establish dynamic load evolution features that characterize the load change trend of urban rail trains. The route environment data acquisition module acquires route environment data based on the route before and after the stops. The route environment data includes gradient data, curvature data, and section environmental disturbance data. The dynamic resistance evolution feature construction module is used to perform section operation resistance correlation analysis based on the line environment data and dynamic load evolution features to establish the dynamic resistance evolution features of the operating section. The dynamic energy consumption constraint field construction module is used to construct the dynamic energy consumption constraint field of operation by utilizing the target operation task and the dynamic resistance evolution characteristics. The dual-path control optimization execution module uses the dynamic energy consumption constraint field and train operation status data to perform dual-path control optimization, which includes kinetic energy recovery optimization before stopping at the station and traction start control optimization after stopping at the station. The train energy consumption management execution module is used to dynamically optimize and manage the energy consumption of urban rail trains based on the optimization results of dual-path control.

Images