Rail locomotive battery cooperative control method based on isomorphic multi-core control
Through the homogeneous multi-core control method, the multi-core processor of the rail locomotive is divided into collaborative optimization, real-time control and safety monitoring core groups, the core groups are dynamically configured and the battery collaborative control strategy is generated. This solves the problems of high cost, poor portability and insufficient real-time performance of traditional heterogeneous multi-core controllers, and realizes efficient and reliable battery collaborative control.
Patent Information
- Application Number
- CN202510915366.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-03
- Publication Date
- 2025-09-12
AI Technical Summary
Traditional heterogeneous multi-core locomotive controllers have high development costs and poor portability, and data synchronization between cores of different architectures is complex, which increases the uncertainty of real-time control and makes it difficult to meet the real-time and reliability requirements of rail transit for battery collaborative control.
A homogeneous multi-core control method is adopted. By dividing the homogeneous multi-core processor into collaborative optimization core group, real-time control core group and safety monitoring core group, the functional core group is dynamically configured, and the battery collaborative control strategy is generated based on the battery cluster state matrix. Safety verification and continuous monitoring are performed through the safety monitoring core group.
It improves the reliability and real-time performance of battery collaborative control, reduces development costs, enhances portability and algorithm iteration adaptability, enables accurate battery status collection and strategy generation, and enhances the intelligent energy management capabilities of rail locomotives.
Smart Images

Figure CN120621151A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of rail transit electronic control, and in particular to a method for coordinated control of rail locomotive batteries based on homogeneous multi-core control. Background Art
[0002] With the advancement of rail transit electronic control technology, the efficiency and reliability of coordinated control of locomotive batteries have become a critical foundation for energy management in new energy locomotives. Currently, traditional heterogeneous multi-core locomotive controllers rely on dedicated hardware acceleration modules, which pose high development costs and poor portability. Furthermore, data synchronization between cores in different architectures is complex, increasing the uncertainty of real-time control. Fixed allocation of hardware resources also makes it difficult to adapt to the demands of algorithm iteration.
[0003] The existing control method relies solely on fixed hardware resource allocation and heterogeneous architecture, resulting in insufficient flexibility in control strategies. This not only increases system development and maintenance costs, but also makes it difficult to meet the stringent requirements of rail transit for the real-time and reliability of battery collaborative control. Summary of the Invention
[0004] In order to solve the above technical problems, the present application provides a railway locomotive battery collaborative control method based on homogeneous multi-core control, which improves the reliability and real-time performance of battery collaborative control, reduces development costs, and improves portability and algorithm iteration adaptability.
[0005] The embodiments of this application disclose the following technical solutions:
[0006] The present application provides a method for coordinated control of railway locomotive batteries based on homogeneous multi-core control, the method comprising:
[0007] determining a plurality of homogeneous multi-core processors of the rail vehicle, and constructing a processor core cluster topology based on the plurality of homogeneous multi-core processors;
[0008] Identifying a current locomotive operating condition of the rail locomotive, and dynamically configuring a plurality of functional core groups based on the current locomotive operating condition and the processor core group topology, the plurality of functional core groups including a collaborative optimization core group, a real-time control core group, and a safety monitoring core group;
[0009] Obtaining a battery cluster state matrix of the rail locomotive, and generating a battery collaborative control strategy based on the battery cluster state matrix and the current locomotive operating condition through the collaborative optimization core group;
[0010] The battery collaborative control strategy is safety-verified by the safety monitoring core group. When the safety verification passes, the real-time control core group performs collaborative control on the battery cluster according to the battery collaborative control strategy, and continuously monitors the execution process of the collaborative control through the safety monitoring core group.
[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0012] This application proposes a method for collaborative control of railway locomotive batteries based on homogeneous multi-core control. By dividing the homogeneous multi-core processor into three core groups: real-time control, collaborative optimization, and safety monitoring, the voltage, current, temperature, and remaining power of each unit in the battery cluster are collected to construct a multi-dimensional state matrix. Based on these parameters, the computing resources of the collaborative optimization core group are configured to achieve accurate battery state collection and strategy generation, ultimately forming an efficient battery and vehicle collaborative control solution. At the same time, combined with the battery state matrix, locomotive operating parameters, and core group resource configuration, power allocation, charging and discharging, and thermal management strategies are scientifically generated, and control parameters are optimized. Through multi-dimensional data fusion and resource dynamic scheduling strategies, the method avoids the high cost and poor real-time performance problems caused by the traditional heterogeneous controller's reliance on hardware acceleration and fixed resource allocation. By utilizing the collaborative optimization of the multi-core architecture and the strategy generator, the problems of high development cost, poor portability, and high uncertainty in real-time control of traditional heterogeneous multi-core controllers are solved, providing a solution for intelligent and precise energy management of railway locomotives.
[0013] The technical solution of this application breaks through the limitations of traditional heterogeneous controllers in cost, real-time performance and flexibility by integrating multi-source data such as battery status parameters, locomotive operating parameters and multi-core resource scheduling parameters, effectively improving the adaptability of battery control and energy management accuracy under complex working conditions, and providing a new path for intelligent energy management of rail locomotives. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0015] Figure 1 A schematic flow chart of a method for coordinated control of railway locomotive batteries based on homogeneous multi-core control provided in an embodiment of the present application;
[0016] Figure 2 A schematic diagram of the process of configuring multiple functional core groups provided in an embodiment of the present application. DETAILED DESCRIPTION
[0017] The present application provides a railway locomotive battery collaborative control method based on homogeneous multi-core control, which is used to solve the technical problems existing in the prior art, such as high development cost and poor portability of heterogeneous multi-core controllers, high uncertainty in real-time control due to complex data synchronization between cores of different architectures, and difficulty in adapting fixed allocation of hardware resources to the requirements of algorithm iteration.
[0018] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.
[0019] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.
[0020] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.
[0021] Example 1, as shown in the attached Figure 1 and attached Figure 2 As shown, the present application provides a railway locomotive battery collaborative control method based on homogeneous multi-core control, the method comprising the following steps:
[0022] S110: Determine a plurality of homogeneous multi-core processors of the rail vehicle, and construct a processor core group topology based on the plurality of homogeneous multi-core processors;
[0023] In the embodiment of the present application, in the railway locomotive battery collaborative control scenario, in order to achieve flexible scheduling and task partitioning of multi-core resources, it is necessary to first clarify the processor hardware architecture and build a topology structure.
[0024] Specifically, the physical number identification of each homogeneous multi-core processor is obtained to form a processor number set. Each processor contains multiple cores, and each core is assigned a unique identifier (such as processor number minus core number) to ensure unique inter-core communication and resource scheduling.
[0025] Furthermore, based on processor number identification and core identifiers, a hierarchical topology construction logic is used to map physical processors and cores into dynamically configurable core cluster nodes, forming a processor core cluster topology. This topology defines the communication paths and resource allocation logic between core clusters, providing the hardware architecture foundation for the subsequent dynamic configuration of functional core clusters.
[0026] Finally, the processor core group topology serves as the underlying framework for resource management, supporting dynamic mapping and load balancing of functional core groups such as real-time control, collaborative optimization, and security monitoring. Relying on the uniqueness of hardware identification and topological hierarchical design, it provides an accurate resource scheduling basis for multi-core collaborative control.
[0027] Step S110 of the method provided in the embodiment of the present application includes:
[0028] Obtaining the numbering identifiers of the multiple homogeneous multi-core processors to obtain multiple processor numbering identifiers, each homogeneous multi-core processor including multiple processor cores, assigning a unique identifier to each processor core, and obtaining multiple processor core identifiers of each homogeneous multi-core processor;
[0029] The processor core group topology is constructed based on the multiple processor number identifiers and multiple processor core identifiers of each of the homogeneous multi-core processors.
[0030] In an embodiment of the present application, in a railway locomotive battery collaborative control scenario, in order to realize dynamic scheduling and task partitioning of multi-core resources, it is necessary to first obtain the unique hardware identification of the processor and construct a topology structure.
[0031] Specifically, the initialization program reads the physical numbers of each homogeneous multi-core processor (e.g., CPU0, CPU1) to form a processor number set. For each core within a processor (e.g., CPU0 includes Core0 to Core7), a unique identifier (e.g., CPU0-Core0) is assigned using the "processor number-core sequence number" format to ensure address uniqueness for inter-core communication and resource scheduling.
[0032] Furthermore, based on the processor numbers and core identifiers, a tree topology construction logic is used to construct a processor core group topology.
[0033] Specifically, the processor number is used as the parent node and the core identifier is used as the child node, which is mapped to a dynamically configurable core cluster node to form a processor core cluster topology including communication paths and resource hierarchies.
[0034] This topology clarifies the data interaction rules between core groups (such as shared memory area access rights and inter-core interrupt priorities), providing a hardware architecture foundation for the subsequent dynamic mapping of functional core groups.
[0035] For example, when the processor contains two homogeneous multi-core chips (CPU0, CPU1), each chip contains 8 cores, 16 unique core identifiers from CPU0-Core0 to CPU1-Core7 are formed through identifier allocation, and then the physical cores are divided into carriers of different core groups (collaborative optimization core group, real-time control core group and security monitoring core group) according to topological rules.
[0036] This topology can dynamically adjust the core group mapping relationship according to the working conditions. For example, when the load is high, the core can be temporarily transferred from one core group to another. Through the uniqueness of hardware identification and dynamic topology configuration, it provides accurate resource scheduling basis for multi-core collaborative control.
[0037] S120: Identifying a current locomotive operating condition of the railway locomotive, and dynamically configuring a plurality of functional core groups based on the current locomotive operating condition and the processor core group topology, wherein the plurality of functional core groups include a collaborative optimization core group, a real-time control core group, and a safety monitoring core group;
[0038] In the embodiment of the present application, in the scenario of coordinated control of railway locomotive batteries, in order to achieve dynamic adaptation of multi-core resources and efficient execution of tasks, it is necessary to complete the core group configuration through working condition identification and topological resource linkage.
[0039] Specifically, the locomotive operating parameters (such as speed, acceleration, battery status) and the rail locomotive model (such as HXD3D) are first collected, and the pre-trained working condition identifier of the model is retrieved through the on-board database.
[0040] Furthermore, the operating condition identifier extracts features from the operating parameters and performs pattern matching against a library of historical operating condition samples, outputting the current locomotive operating condition type and parameters. Based on the current operating condition characteristics, the functional core group performance constraints are determined, which must meet response time, computational complexity, and safety redundancy requirements.
[0041] Furthermore, a core group configuration space is constructed based on the processor core group topology, and the number constraints of each core group and the total number constraints are set. By introducing time fitness, complexity fitness, and safety fitness functions, multiple sets of initial core group configuration schemes are randomly generated within the configuration space.
[0042] Finally, multiple initial core group configuration schemes are iteratively optimized based on the time fitness function, complexity fitness function and safety fitness function. When the time fitness of a scheme meets the response time requirements, the complexity fitness meets the computational complexity requirements and the safety fitness meets the safety redundancy requirements, the target core group configuration scheme is output to generate collaborative optimization core groups, real-time control core groups and safety monitoring core groups to achieve precise matching of hardware resources and working conditions.
[0043] Step S120 in the method provided in the embodiment of the present application includes:
[0044] Obtaining locomotive operating parameters and a locomotive model of the locomotive, and retrieving an operating condition identifier corresponding to the locomotive model;
[0045] Performing operating condition identification processing on the locomotive operating parameters by the operating condition identifier to obtain the current locomotive operating condition;
[0046] determining a functional core group performance constraint according to the current locomotive operating condition, wherein the functional core group performance constraint includes a response time requirement, a computational complexity requirement, and a safety redundancy requirement;
[0047] Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, the collaborative optimization core group, the real-time control core group, and the safety monitoring core group are configured in combination with the processor core group topology.
[0048] In the embodiment of the present application, in the railway locomotive battery collaborative control scenario, in order to achieve accurate adaptation of multi-core resources and operating conditions, it is necessary to first establish a basis for hardware resource scheduling by building a condition identifier.
[0049] The steps of constructing the “operating condition identifier” in the method provided in the embodiment of the present application include:
[0050] Extracting historical locomotive operating parameters of locomotives of the same model according to the locomotive model, constructing a sample locomotive operating parameter set, and labeling the sample locomotive operating parameter set with operating conditions to obtain a sample locomotive operating condition label set;
[0051] The operating condition identifier is obtained by training based on the sample locomotive operating parameter set and the sample locomotive operating condition label set using machine learning.
[0052] In the embodiment of the present application, in order to improve the adaptability and classification accuracy of operating condition identification to complex operating scenarios, it is necessary to construct an operating condition identifier to achieve accurate mapping from locomotive operating parameters to operating condition types.
[0053] First, for the target rail locomotive model (such as HXD3D), locomotive operating parameters covering different operating stages, load conditions and environmental conditions are extracted from the historical operation database, including multi-dimensional data such as traction motor speed, battery pack voltage, braking current, ambient temperature, ramp gradient, etc. After data cleaning (eliminating outliers and filling missing values) and feature standardization, a structured sample locomotive operating parameter set is formed.
[0054] This dataset covers typical operating conditions such as starting acceleration, constant speed driving, braking deceleration, climbing, and descending, as well as special scenarios such as low temperature and high load.
[0055] On this basis, professionals label the operating condition type for each set of data in the sample parameter set according to the correspondence between the operating parameter characteristics and the actual operating conditions, forming a sample locomotive operating condition label set that includes operating condition labels (such as "low temperature start", "climbing heavy load", "brake energy recovery", etc.).
[0056] Furthermore, after completing the construction of the sample locomotive operating parameter set and the sample locomotive operating condition label set, the support vector machine (SVM) algorithm is used as the basic architecture to construct a working condition identifier to solve the problem of insufficient adaptability of traditional identification methods in multi-dimensional parameter coupling scenarios.
[0057] Specifically, the SVM algorithm was adopted as the core architecture, with the radial basis function (RBF) chosen as the kernel function to balance the model's nonlinear mapping capabilities and computational efficiency. During training, key parameters were optimized using a grid search method. The search range for the penalty coefficient C was set to [0.1, 100], and the RBF kernel function parameter γ was set to [0.01, 10]. Minimizing the 10-fold cross-validation error was the goal. The optimal parameter combination (C = 10, γ = 0.1) was determined based on the parameter characteristics of new energy rail locomotives under typical operating conditions, such as starting at low temperatures and climbing with heavy loads.
[0058] For example, in the historical data verification of the HXD3D locomotive, the recognition error rate of this parameter combination for the "low-temperature start-up" operating condition was reduced to 3.2%, a significant improvement over the initial parameters.
[0059] The labeled sample set was further divided into a training set (1,600 groups) and a test set (400 groups) in an 8:2 ratio and fed into the SVM model for iterative training. The model's input layer receives the standardized parameter vector, maps the data into a high-dimensional feature space using a kernel function, and constructs a decision hyperplane using the maximum margin classification principle.
[0060] During the training phase, an early stopping strategy is used to avoid overfitting. That is, when the classification accuracy on the test set increases by less than 0.5% for five consecutive rounds, the training is automatically terminated and the model is judged to have converged.
[0061] At the same time, in order to enhance the model's adaptability to sensor errors and sudden changes in working conditions, data augmentation operations are performed on the training set. That is, by adding ±5°C random perturbations to the temperature parameters and scaling the load current by ±15%, 400 groups of derivative samples are generated, expanding the total amount of training data to 2400 groups.
[0062] In the end, the trained operating condition identifier achieved a classification accuracy of 96% on the test set, and the recall rate for complex operating conditions such as "low-temperature start-up" and "brake energy recovery" reached over 93%.
[0063] The identifier can receive the locomotive operating parameter stream in real time. Through the high-dimensional feature mapping and rapid decision-making capabilities of SVM, it can complete the current operating condition type determination in a short time, providing a reliable basis for the dynamic configuration of subsequent functional core groups.
[0064] After the working condition identifier is trained, the locomotive operating parameters and rail locomotive model collected in real time are input into the working condition identifier. The parameters are subjected to feature extraction and high-dimensional space mapping through the trained SVM model, and the constructed decision hyperplane is used to complete the working condition classification and obtain the current working condition of the locomotive.
[0065] Furthermore, based on the identified current locomotive operating conditions, the performance constraints of the functional core groups are determined. These constraints include response time requirements, computational complexity requirements, and safety redundancy requirements. This ensures that the subsequently configured collaborative optimization core group, real-time control core group, and safety monitoring core group can meet these three requirements when processing tasks.
[0066] For example, under low-temperature starting conditions, the battery heating strategy calculation must be completed within 15ms (response time requirement), involving 300 iterative operations of the thermal coupling matrix (computational complexity requirement), and dual-core cross-checking must be enabled (safety redundancy requirement); under braking energy recovery conditions, braking torque control must be guaranteed within a 2ms cycle (response time requirement), and the computational complexity is 100 floating-point operations per second of the motor model, and safety redundancy must meet single-fault tolerance.
[0067] Furthermore, based on the response time requirements, the computational complexity requirements and the safety redundancy requirements, the core group, the real-time control core group and the safety monitoring core group are collaboratively optimized in combination with the processor core group topology configuration.
[0068] In the method provided in an embodiment of the present application, the step of “configuring the collaborative optimization core group, the real-time control core group, and the security monitoring core group in combination with the processor core group topology based on the response time requirement, the computational complexity requirement, and the security redundancy requirement” includes:
[0069] Setting core group quantity constraints for the collaborative optimization core group, the real-time control core group, and the security monitoring core group according to the processor core group topology to obtain a first core group quantity constraint, a second core group quantity constraint, a third core group quantity constraint, and a core group total quantity constraint;
[0070] Constructing a core group configuration space according to the first core group number constraint, the second core group number constraint, the third core group number constraint, and the core group total amount constraint;
[0071] Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, a core group configuration optimization is performed in the core group configuration space to determine the collaborative optimization core group, the real-time control core group, and the safety monitoring core group.
[0072] In the embodiment of the present application, in order to achieve an accurate match between hardware resources and working condition requirements, it is necessary to first set constraints on the number of three functional core groups based on the hardware architecture characteristics of the processor core group topology.
[0073] Specifically, the hardware foundation of the processor core cluster topology is first defined. This topology consists of the numbering, core identifiers, and communication paths of the homogeneous multi-core processors. It also defines the total number of core resources (e.g., 16 cores) and their allocation rules (e.g., prioritizing cores within the same chip).
[0074] Furthermore, boundary constraints are set based on the functional characteristics of the core group.
[0075] Among them, the collaborative optimization core group (the first core group) needs to rely on parallel computing capabilities to support matrix operation acceleration strategies for complex computing tasks such as battery thermal management and power allocation. Therefore, the core number range is set based on the total number of topological cores, for example, 4 to 8 cores.
[0076] The real-time control core group (second core group) targets sub-millisecond real-time tasks such as motor control. Based on the core affinity rules in the topology (lower communication latency between cores within the same chip), the number of cores is constrained to 3 to 6, and tasks are fixed to specific cores through CPU affinity binding technology to ensure the execution determinism of tasks such as braking torque control.
[0077] The security monitoring core group (the third core group) takes into account the cross-layer security protection requirements, such as implementing dual-core verification and conducting fault monitoring. The number of cores is set to 2 to 4 based on the topology fault tolerance mechanism, thereby ensuring cross-verification of output results of different core groups and realizing dual-machine redundancy of shared memory area verification.
[0078] Finally, establish a total core cluster constraint. Based on the upper limit of the total core count of the processor core cluster topology (e.g., 16), stipulate that the sum of the core counts of the three core clusters must not exceed this value to avoid resource allocation conflicts. For example, with 8 collaborative optimization core clusters, 6 real-time control core clusters, and 2 security monitoring core clusters, a total of 16 cores meets the constraint and ensures a stable hardware architecture.
[0079] Furthermore, a core group configuration space is constructed through the above quantity constraints. This configuration space is bounded by the quantity constraints of the collaborative optimization core group, the real-time control core group, and the safety monitoring core group, forming a set of all feasible core allocation combinations.
[0080] For example, when the first core group quantity constraint of the collaborative optimization core group is 4 to 8 cores, the second core group quantity constraint of the real-time control core group is 3 to 6 cores, the third core group quantity constraint of the security monitoring core group is 2 to 4 cores, and the total number of core groups is constrained not to exceed 16 cores, the core group configuration space includes all core quantity combinations that meet the above interval restrictions, such as (7, 3, 2), (5, 4, 3), (6, 5, 3), etc.
[0081] Furthermore, according to known response time requirements, computational complexity requirements, and safety redundancy requirements, a collaborative optimization core group, a real-time control core group, and a safety monitoring core group are determined through iterative optimization in the core group configuration space.
[0082] In the method provided in an embodiment of the present application, the step of “performing a core group configuration optimization in the core group configuration space based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, and determining the collaborative optimization core group, the real-time control core group, and the safety monitoring core group” includes:
[0083] Obtaining time fitness function, complexity fitness function and security fitness function;
[0084] Based on the first core group number constraint, the second core group number constraint, the third core group number constraint, and the core group total amount constraint, randomly acquiring a plurality of initial core group configuration schemes in the core group configuration space;
[0085] Obtaining time fitness, complexity fitness, and safety fitness of the multiple initial core group configuration schemes according to the time fitness function, the complexity fitness function, and the safety fitness function, and iteratively optimizing the multiple initial core group configuration schemes;
[0086] When the time fitness meets the response time requirement, the complexity fitness meets the computational complexity requirement, and the safety fitness meets the safety redundancy requirement in the optimization result, the target core group configuration scheme is output to obtain the collaborative optimization core group, the real-time control core group, and the safety monitoring core group.
[0087] In the embodiment of the present application, in order to achieve accurate matching of hardware resources and working condition requirements, it is necessary to perform core group configuration optimization in the core group configuration space to determine the collaborative optimization core group, real-time control core group and safety monitoring core group.
[0088] Specifically, firstly, by analyzing the mapping relationship between system performance requirements and core group resource configuration under different working conditions, the time fitness function, complexity fitness function and safety fitness function are obtained.
[0089] Among them, the time fitness function is obtained by sorting out the real-time requirements under typical operating conditions of rail transit (such as low-temperature start-up, brake energy recovery, etc.), and constructing the function based on the maximum allowable response time of key tasks.
[0090] For example, during low-temperature startup, the collaborative optimization core group needs to complete the battery heating strategy calculation within 15ms. If the calculation time of a certain solution is ≤15ms, the time adaptability value increases as time shortens. Otherwise, it is judged as not meeting the real-time requirements. This function is used to measure the degree to which the core group configuration solution meets the system response time. The shorter the response time (such as the critical task is completed within 15ms), the higher the function value, reflecting the configuration's adaptability to real-time requirements.
[0091] The complexity fitness function is derived by building a function based on the matching efficiency between the computational workload (e.g., number of matrix operations, floating-point operation frequency) and the number of cores in the core group. This function reflects the efficiency of computing resource allocation; a higher ratio indicates more efficient resource utilization.
[0092] The safety fitness function is constructed based on the redundancy mechanisms of the security monitoring core group and the system's fault handling capabilities. This function is used to correlate the redundancy capabilities of the security monitoring core group (such as the dual-core verification mechanism) with reliability. This function quantifies the safety level. Higher redundancy and lower failure probability increase the function value, ensuring that the configuration meets safety protection requirements.
[0093] Furthermore, based on the first core group quantity constraint (such as 4-8 cores in the collaborative optimization core group), the second core group quantity constraint (3-6 cores in the real-time control core group), the third core group quantity constraint (2-4 cores in the safety monitoring core group) and the total number of core groups constraint (such as the total number of cores ≤ 16), multiple initial core group configuration schemes are randomly obtained in the core group configuration space, such as core quantity combinations such as (5, 4, 3) and (7, 3, 2).
[0094] Furthermore, the time fitness, complexity fitness and safety fitness of each initial solution are calculated through the above three types of fitness functions, and the solution is optimized using an iterative optimization algorithm (such as a genetic algorithm). In each iteration, the solution with higher fitness is retained and the configuration that does not meet the requirements is eliminated.
[0095] Specifically, in the iterative optimization process, taking the genetic algorithm as an example, the initial core group configuration scheme (such as 5 cores in the collaborative optimization core group, 4 cores in the real-time control core group, and 3 cores in the security monitoring core group) is first encoded, and a new generation of schemes is generated through operations such as crossover and mutation.
[0096] In each iteration, all solutions are evaluated using three types of fitness functions, and solutions whose time fitness meets the response time requirements, complexity fitness meets the computing efficiency requirements, and security fitness meets the security redundancy standards are selected as parents, and configurations that do not meet the requirements of any dimension are eliminated.
[0097] Furthermore, through multiple iterations, the core count allocation for each core group was gradually adjusted. For example, in one iteration, it was found that increasing the number of cores in the collaborative optimization core group from 5 to 6 cores reduced the battery thermal management calculation time from 18ms to 14ms, meeting the 15ms response time requirement. This adjustment direction was then retained, while verifying that the core counts in the real-time control core group and the safety monitoring core group still met their respective complexity and safety requirements.
[0098] Ultimately, through continuous optimization, the fitness of the solutions in the optimization results in all three dimensions reaches the preset threshold, and the target core group configuration plan is output (such as 6 cores in the collaborative optimization core group, 5 cores in the real-time control core group, and 3 cores in the safety monitoring core group), achieving a high degree of fit between the precise configuration of hardware resources and the working conditions.
[0099] S130: Obtaining a battery cluster state matrix of the rail vehicle, and generating a battery collaborative control strategy based on the battery cluster state matrix and the current locomotive operating condition through the collaborative optimization core group;
[0100] In the embodiment of the present application, in the scenario of coordinated control of railway locomotive batteries, in order to achieve efficient management of battery energy and dynamic adaptation of working conditions, it is necessary to generate a control strategy through battery status collection and coordinated optimization of core group operations.
[0101] Specifically, the voltage, current, temperature and remaining power of each battery cell in the rail locomotive battery cluster are first collected through sensors, and a multi-dimensional battery cluster state matrix is constructed according to a preset format.
[0102] The matrix uses battery cells as row dimensions and state parameters such as voltage, current, temperature, and remaining power as column dimensions to map the physical state of the battery cluster in real time.
[0103] Furthermore, the pre-trained battery control strategy generator is retrieved and initialized with the current configuration of the collaborative optimization core group. The battery cluster state matrix and the current locomotive operating condition information are synchronously input into the generator. The collaborative optimization core group's multiple processor cores perform matrix operations in parallel, gradually calculating the battery power allocation plan, charging and discharging strategy, and thermal management strategy.
[0104] Finally, based on the collaborative optimization results of the three types of strategies, a battery collaborative control strategy is generated to achieve real-time collaborative management of the battery cluster, ensuring efficient utilization of battery energy and safe operation under different working conditions.
[0105] Step S130 in the method provided in the embodiment of the present application includes:
[0106] Collect the voltage, current, temperature and remaining capacity of each battery cell in the railway locomotive battery cluster to build a battery cluster status matrix;
[0107] Retrieving a battery control strategy generator and configuring the battery control strategy generator with the collaborative optimization core group, inputting the battery cluster state matrix and the current locomotive operating condition into the battery control strategy generator, and using multiple processor cores in the collaborative optimization core group to parallelly calculate a battery power allocation plan, a charge and discharge strategy, and a thermal management strategy;
[0108] A battery collaborative control strategy is generated based on the battery power allocation scheme, the charge and discharge strategy, and the thermal management strategy.
[0109] In the battery energy management scenario of a rail locomotive in the embodiment of the present application, in order to achieve the accuracy and real-time performance of the coordinated control of the battery and the vehicle, it is necessary to generate a complete battery control solution through multi-dimensional state acquisition, matrix construction, parallel computing and strategy collaboration.
[0110] Specifically, first, the voltage, current, temperature and remaining power of each battery cell are collected in real time through sensors distributed in the battery cluster (including closed-loop Hall effect sensors, current sensors and NTC thermistors), and a battery cluster status matrix is constructed in a preset format with battery cells as row dimensions and voltage / current / temperature / remaining power as column dimensions to map the physical state of the battery cluster in real time.
[0111] Among them, voltage and current data are used to evaluate the battery's charging and discharging capabilities, temperature data is related to thermal management requirements (such as heat dissipation at high temperatures and heating at low temperatures), and the remaining power reflects the remaining energy level of the battery.
[0112] Furthermore, a pre-trained battery control strategy generator is retrieved from the controller software library and initialized with the current configuration of the collaborative optimization kernel group.
[0113] The steps for constructing the "battery control strategy generator" in the method provided in the embodiment of the present application include:
[0114] Collecting historical operating data of the battery cluster under different locomotive operating conditions, wherein the historical operating data of the battery cluster includes a sample battery cluster state matrix and sample locomotive operating condition information;
[0115] Annotating an optimal control strategy according to the sample battery cluster state matrix and the sample locomotive operating condition information to obtain a sample battery collaborative control strategy;
[0116] The battery control strategy generator is generated through supervised learning training, using the sample battery cluster state matrix and the sample locomotive operating condition information as input features and the sample battery collaborative control strategy as a supervisory signal.
[0117] In an embodiment of the present application, in order to improve the adaptability of the battery control strategy generator to complex working conditions and the accuracy of strategy generation, a hybrid algorithm framework that integrates incremental learning and transfer learning is adopted to construct a battery control strategy generator to form a strategy generation capability covering multiple scenarios.
[0118] Specifically, the historical operation data of the battery cluster under different locomotive operating conditions such as traction, braking, and standby are first collected, including the sample battery cluster state matrix (parameters such as each battery cell voltage, current, temperature, remaining power, etc., sampling frequency ≥ 100Hz) and sample locomotive operating condition information (locomotive speed, acceleration, power demand, track slope, etc.).
[0119] At the same time, the original data is denoised through Kalman filtering, and classified and sorted according to the operating condition type (such as low-temperature starting, high-speed traction, and emergency braking) to construct a structured training data set covering multiple typical scenarios.
[0120] On this basis, we collaborated with battery management experts and locomotive control engineers to label the optimal control strategies for sample data under different working conditions, including power allocation ratio, thermal management logic, safety protection thresholds, etc., and converted them into numerical sample battery collaborative control strategies (such as "battery heating power is 500W at -20℃, and batteries with remaining power ≥ 60% have priority output power").
[0121] Furthermore, after dividing the dataset into training set, validation set and test set at the ratio of 8:1:1, the hybrid architecture of "graph neural network (GNN) combined with long short-term memory network (LSTM)" was used for model training to construct a battery control strategy generator.
[0122] Among them, the GNN layer processes the relationship between battery cells through matrix sparse operations (only retaining the temperature-related parameters of the main nodes to construct a sparse matrix), and the LSTM layer performs matrix transformation on the time series of operating parameters to capture dynamic characteristics.
[0123] During the training process, an incremental learning strategy is adopted, that is, 128 samples are extracted each time for small-batch gradient descent, and transfer learning is introduced, using the low-temperature working condition training parameters as the high-temperature working condition initialization parameters to accelerate model convergence.
[0124] At the same time, an early stopping mechanism is set to terminate training when the validation set loss decreases by less than 0.5% for 10 consecutive rounds to avoid overfitting and determine whether the model is fitted.
[0125] After training is completed, the model performance is evaluated through the test set, requiring that the power distribution strategy error under traction conditions is less than 5%, the charging control accuracy under braking conditions is greater than 95%, and the response time of the safety strategy under abnormal conditions is less than 100ms.
[0126] At the same time, knowledge distillation technology is used to compress model parameters (remove redundant parameters), and matrix operations are used to ensure that the calculation time on the collaborative optimization core group (4-8 cores) is less than 15ms. Finally, the optimized generator is integrated into the locomotive controller software to realize the dynamic generation of collaborative control strategies based on real-time battery status and operating conditions.
[0127] Through multi-dimensional data training and hybrid network architecture, the generator avoids the policy deviation of a single model under complex working conditions and effectively improves the accuracy of policy generation in various scenarios.
[0128] Furthermore, when a real-time control strategy needs to be generated, the battery control strategy generator is called and configured with the collaborative optimization core cluster. The current battery cluster state matrix and locomotive operating conditions are input. Through parallel computation across multiple processor cores in the collaborative optimization core cluster, numerical results are output for the power allocation plan, charge-discharge strategy, and thermal management strategy.
[0129] Specifically, in terms of power allocation scheme generation, a lightweight MPC algorithm is used to perform matrix operations based on the locomotive's real-time power demand matrix and the battery status matrix to calculate the power allocation ratio of each battery unit.
[0130] For example, when 2000kW is required to be output under traction conditions, the generator uses matrix multiplication operations to prioritize battery cells with a charge state of 60% to 80% to undertake the main power output, and stores the allocation results as a power allocation matrix to avoid overloading of a single battery group.
[0131] When generating the charge and discharge strategy, the battery remaining capacity threshold matrix and the working condition energy flow matrix are combined to generate the charge and discharge logic control matrix.
[0132] For example, when the remaining battery power is ≥95%, the current limiting protection matrix instruction is triggered, the regenerative braking energy is switched to resistance braking, and the charging current upper limit threshold is updated through matrix operation; if the remaining battery power is <90%, the regenerative braking charging matrix is enabled and the charging current threshold is set to 200A.
[0133] During the thermal management strategy generation process, the cooling / heating control matrix is output based on the battery temperature distribution matrix and the thermal coupling model matrix.
[0134] For example, if the temperature of a battery pack is greater than 40°C, a matrix instruction is generated to start the fan to dissipate heat. When the temperature is less than 0°C, the heating film is activated and the power distribution matrix is adjusted to assist in heating by discharging some batteries and generating heat. At the same time, the thermal management matrix and the power distribution matrix are coordinated to prevent further temperature increases.
[0135] Finally, the generator performs a tensor fusion operation on the power allocation matrix, the charge-discharge logic matrix, and the thermal management matrix to generate a multidimensional battery collaborative control strategy matrix. After the activation function of the fully connected layer, this matrix is converted into executable control instructions, which serve as the battery collaborative control strategy.
[0136] For example, under low-temperature starting conditions, the strategy matrix integrates the "heating power 500W matrix + battery remaining power ≥ 60%, battery priority output matrix" and generates a composite strategy including heating control and power distribution through tensor operations; under braking conditions, the "regenerative braking charging current 200A matrix" and the "battery remaining power ≥ 95% switching resistor braking matrix" are integrated, and after activation function processing, a battery collaborative control strategy that combines charging control and safe switching is formed.
[0137] S140: The battery collaborative control strategy is safety-verified by the safety monitoring core group. When the safety verification passes, the real-time control core group performs collaborative control on the battery cluster according to the battery collaborative control strategy, and continuously monitors the execution process of the collaborative control by the safety monitoring core group.
[0138] In the embodiment of the present application, in the scenario of coordinated control of railway locomotive batteries, in order to achieve safe execution of control strategies and reliable monitoring of the entire process, strategy verification and closed-loop control need to be implemented through multi-layer security assurance.
[0139] Specifically, the safety monitoring core group first conducts safety boundary verification and feasibility assessment on the generated battery collaborative control strategy. By comparing historical safety parameters such as the safety threshold range of battery cell voltage / current and temperature protection limits, it verifies whether the power allocation scheme in the strategy poses overload risks and whether the charging and discharging logic exceeds safety boundaries.
[0140] For example, when it is detected that the discharge current of a battery pack exceeds the safety threshold, the policy degradation mechanism is automatically triggered and replaced with a historically verified safety parameter combination.
[0141] If the verification passes, the real-time control core group sends control instructions to the battery management system and motor controller through a dedicated communication channel to perform operations such as power distribution and charge and discharge regulation.
[0142] At the same time, the safety monitoring core group collects the real-time operating data of the battery cluster (voltage fluctuations, temperature changes, current output, etc.) at a sub-millisecond frequency, and dynamically compares it with the control targets preset by the strategy to form a closed-loop monitoring link.
[0143] If the parameter deviation exceeds the preset safety threshold for three consecutive times, the system safety response mechanism will be triggered immediately, the abnormal control command will be cut off and the fault information will be reported to the on-board monitoring system.
[0144] During the collaborative control process, the safety monitoring core cluster uses independent hardware monitoring circuits to monitor the operating status of each core cluster in real time. If the task response time of the real-time control core cluster or the collaborative optimization core cluster exceeds the millisecond threshold, the backup core resource will be automatically activated to take over the control task, ensuring the timely execution of critical operations such as braking energy recovery and emergency power off.
[0145] This mechanism achieves a full-process safety closed loop from strategy generation to execution through multiple steps of strategy security verification, execution process monitoring, and hardware redundancy protection, meeting the stringent requirements for safety and reliability in the rail transit field.
[0146] The embodiments of the present application achieve the following technical effects through the above specific implementation methods:
[0147] The embodiment of the present application provides a method for collaborative control of railway locomotive batteries based on homogeneous multi-core control. First, multiple homogeneous multi-core processors of the railway locomotive are determined, the number identification of each processor and the unique identifier of each core are obtained, and a hierarchical processor core group topology is constructed to lay the foundation for flexible scheduling of multi-core resources and task partitioning. Secondly, the locomotive operating parameters are collected and the corresponding model working condition identifier is called. After identifying the current working condition, the performance constraints of the functional core group in terms of response time, computational complexity and safety redundancy are determined. Then, based on the topology, the number constraints of the core group are set, and the configuration space is constructed. The fitness function is used to iteratively search for the best, and three types of core groups, namely collaborative optimization, real-time control and safety monitoring, are dynamically configured to achieve precise matching of hardware resources and working condition requirements. Next, the battery status is collected to construct a matrix, and the pre-training generator is called. The power allocation, charging and discharging and thermal management strategies are calculated in parallel by the collaborative optimization core group, and the collaborative control strategy is generated by tensor fusion. Finally, after the safety monitoring core group verification strategy passes, the real-time control core group executes the instructions and continuously monitors to ensure that the control process is safe and reliable.
[0148] The method provided in the embodiment of the present application solves the problems of high cost and poor real-time performance of traditional heterogeneous controllers. By determining the processor construction topology, identifying the working condition configuration functional core group, etc., it realizes the coordinated control of the battery and the vehicle, improves the adaptability to complex working conditions and the energy management accuracy, and provides an innovative path for the intelligent energy management of rail locomotives.
[0149] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0150] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.
[0151] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.
Claims
1. A railway locomotive battery collaborative control method based on homogeneous multi-core control, characterized in that: The method comprises: determining a plurality of homogeneous multi-core processors of the rail vehicle, and constructing a processor core cluster topology based on the plurality of homogeneous multi-core processors; Identifying a current locomotive operating condition of the rail locomotive, and dynamically configuring a plurality of functional core groups based on the current locomotive operating condition and the processor core group topology, the plurality of functional core groups including a collaborative optimization core group, a real-time control core group, and a safety monitoring core group; Obtaining a battery cluster state matrix of the rail locomotive, and generating a battery collaborative control strategy based on the battery cluster state matrix and the current locomotive operating condition through the collaborative optimization core group; The battery collaborative control strategy is safety-verified by the safety monitoring core group. When the safety verification passes, the real-time control core group performs collaborative control on the battery cluster according to the battery collaborative control strategy, and continuously monitors the execution process of the collaborative control through the safety monitoring core group.
2. The method according to claim 1, characterized in that Building a processor core group topology based on the multiple homogeneous multi-core processors includes: Obtaining the numbering identifiers of the multiple homogeneous multi-core processors to obtain multiple processor numbering identifiers, each homogeneous multi-core processor including multiple processor cores, assigning a unique identifier to each processor core, and obtaining multiple processor core identifiers of each homogeneous multi-core processor; The processor core group topology is constructed based on the multiple processor number identifiers and multiple processor core identifiers of each of the homogeneous multi-core processors.
3. The method according to claim 1, characterized in that Identifying a current locomotive operating condition of the rail locomotive, and dynamically configuring a plurality of functional core groups based on the current locomotive operating condition and the processor core group topology, the plurality of functional core groups including a collaborative optimization core group, a real-time control core group, and a safety monitoring core group, including: Obtaining locomotive operating parameters and a locomotive model of the locomotive, and retrieving an operating condition identifier corresponding to the locomotive model; Performing operating condition identification processing on the locomotive operating parameters by the operating condition identifier to obtain the current locomotive operating condition; determining a functional core group performance constraint according to the current locomotive operating condition, wherein the functional core group performance constraint includes a response time requirement, a computational complexity requirement, and a safety redundancy requirement; Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, the collaborative optimization core group, the real-time control core group, and the safety monitoring core group are configured in combination with the processor core group topology.
4. The method according to claim 3, characterized in that The steps of constructing the operating condition identifier include: Extracting historical locomotive operating parameters of locomotives of the same model according to the locomotive model, constructing a sample locomotive operating parameter set, and labeling the sample locomotive operating parameter set with operating conditions to obtain a sample locomotive operating condition label set; The operating condition identifier is obtained by training based on the sample locomotive operating parameter set and the sample locomotive operating condition label set using machine learning.
5. The method according to claim 3, characterized in that Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, and in combination with the processor core group topology, configuring the collaborative optimization core group, the real-time control core group, and the safety monitoring core group includes: Setting core group quantity constraints for the collaborative optimization core group, the real-time control core group, and the security monitoring core group according to the processor core group topology to obtain a first core group quantity constraint, a second core group quantity constraint, a third core group quantity constraint, and a core group total quantity constraint; Constructing a core group configuration space according to the first core group number constraint, the second core group number constraint, the third core group number constraint, and the core group total amount constraint; Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, a core group configuration optimization is performed in the core group configuration space to determine the collaborative optimization core group, the real-time control core group, and the safety monitoring core group.
6. The method according to claim 5, characterized in that Based on the response time requirement, the computational complexity requirement, and the safety redundancy requirement, performing core group configuration optimization in the core group configuration space to determine the collaborative optimization core group, the real-time control core group, and the safety monitoring core group, including: Obtaining time fitness function, complexity fitness function and security fitness function; Based on the first core group number constraint, the second core group number constraint, the third core group number constraint, and the core group total amount constraint, randomly acquiring a plurality of initial core group configuration schemes in the core group configuration space; Obtaining time fitness, complexity fitness, and safety fitness of the multiple initial core group configuration schemes according to the time fitness function, the complexity fitness function, and the safety fitness function, and iteratively optimizing the multiple initial core group configuration schemes; When the time fitness meets the response time requirement, the complexity fitness meets the computational complexity requirement, and the safety fitness meets the safety redundancy requirement in the optimization result, the target core group configuration scheme is output to obtain the collaborative optimization core group, the real-time control core group, and the safety monitoring core group.
7. The method according to claim 1, characterized in that Obtaining a battery cluster state matrix of the rail locomotive, and generating a battery collaborative control strategy based on the battery cluster state matrix and the current locomotive operating condition through the collaborative optimization core group, including: Collect the voltage, current, temperature and remaining capacity of each battery cell in the railway locomotive battery cluster to build a battery cluster status matrix; Retrieving a battery control strategy generator and configuring the battery control strategy generator with the collaborative optimization core group, inputting the battery cluster state matrix and the current locomotive operating condition into the battery control strategy generator, and using multiple processor cores in the collaborative optimization core group to parallelly calculate a battery power allocation plan, a charge and discharge strategy, and a thermal management strategy; A battery collaborative control strategy is generated based on the battery power allocation scheme, the charge and discharge strategy, and the thermal management strategy.
8. The method according to claim 7, characterized in that The steps of constructing the battery control strategy generator include: Collecting historical operating data of the battery cluster under different locomotive operating conditions, wherein the historical operating data of the battery cluster includes a sample battery cluster state matrix and sample locomotive operating condition information; Annotating an optimal control strategy according to the sample battery cluster state matrix and the sample locomotive operating condition information to obtain a sample battery collaborative control strategy; The battery control strategy generator is generated through supervised learning training, using the sample battery cluster state matrix and the sample locomotive operating condition information as input features and the sample battery collaborative control strategy as a supervisory signal.
