Electric bus double-motor cooperative control method, system, device and medium

Abstract

This invention relates to the field of electric bus control, and particularly to a method, system, device, and medium for dual-motor coordinated control of an electric bus. The method involves acquiring multi-source state data of a pure electric bus during operation and constructing a working condition feature vector. Based on the working condition feature vector, historical operating data is subjected to working condition clustering analysis to establish a mapping relationship between working condition features and coordinated control parameters. A dual-motor coordinated control parameter optimization model is constructed, and under constraints of motor efficiency spectrum, battery power boundary, and system stability, the coordinated control parameters are optimized through multi-objective optimization to obtain the optimal coordinated control parameter vector. This enhances the adaptability of the vehicle control system under complex operating conditions, extends the service life of the motor and power system, improves the stability and reliability of the vehicle control system, reduces the impact of torque fluctuations, and improves passenger comfort.

Description

Technical Field

[0001] This invention relates to the field of electric bus control, and in particular to a method, system, device and medium for coordinated control of dual motors in an electric bus. Background Technology

[0002] With the widespread application of pure electric buses in public transportation, the vehicle control system needs to simultaneously consider power performance, economy, passenger comfort, and system reliability under varying operating conditions. Compared to the traditional single-motor drive structure, the dual-motor drive system, through torque coupling and operating mode switching between the two drive motors, can achieve a wider efficient operating range and more flexible drive strategies at the vehicle control level, and is therefore widely used in large electric buses.

[0003] In existing technologies, the overall control system of electric buses typically controls the torque distribution and mode switching of the dual motors through preset control parameters. However, in actual operation, vehicle operating conditions (such as starting, acceleration, cruising, and climbing) change frequently, and different operating conditions have significantly different requirements for control response speed, shift smoothness, motor efficiency, and system lifespan. Furthermore, existing control methods still have the following shortcomings in parameter optimization and application:

[0004] First, in the control parameter optimization stage, a random initialization method with fixed upper and lower bounds is often used without taking into account the actual operating conditions of the electric bus and the vehicle constraints such as motor efficiency graphs and battery power boundaries. This results in a large number of the generated initial control parameters being in infeasible or inefficient regions, affecting the optimization efficiency and convergence performance of the vehicle control system.

[0005] Secondly, under the requirement of multi-objective control, existing methods usually transform multi-objective problems into single-objective optimization with fixed weights, which makes it difficult to reflect the differentiated requirements of vehicle control strategies under different operating conditions. This makes it difficult for dual motors to simultaneously take into account power response, smoothness and energy efficiency in actual operation.

[0006] Furthermore, in terms of optimization algorithms, existing gray wolf optimization algorithms mostly use fixed convergence factors, which cannot dynamically adjust the search strategy according to the complexity of the current operating conditions of the electric bus. At the same time, they lack guidance on motor efficiency characteristics and torque response constraints during parameter updates, which can easily lead to parameter solutions that do not meet the overall vehicle control requirements.

[0007] Furthermore, existing technologies typically obtain fixed control parameters through a single offline optimization, without establishing a matching relationship between operating conditions and control parameters, and lacking an assessment of the stability of parameters under different operating conditions. This makes it difficult for the optimization results to adapt to the changes in operating conditions of electric buses in actual operation, thereby affecting the collaborative control effect of dual motors and the overall vehicle performance.

[0008] Therefore, it is necessary to propose a collaborative control method for the dual-motor drive control system of electric buses that can combine multi-condition characteristics and achieve adaptive parameter optimization and matching. Summary of the Invention

[0009] The purpose of this invention is to solve the problems mentioned in the background section.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A method for coordinated control of dual motors in an electric bus, the method comprising:

[0012] Acquire multi-source state data of pure electric buses during operation and construct operating condition feature vectors;

[0013] Based on the operating condition feature vector, perform operating condition clustering analysis on historical operating data to construct a mapping relationship between operating condition features and collaborative control parameters;

[0014] A dual-motor cooperative control parameter optimization model is constructed, and the cooperative control parameters are optimized under the constraints of motor efficiency spectrum, battery power boundary and system stability to obtain the optimal cooperative control parameter vector;

[0015] Establish a parameter-operating condition matching library, and identify the operating condition based on the current operating condition feature vector, and retrieve the corresponding optimal cooperative control parameter vector from the matching library;

[0016] Based on the optimal cooperative control parameter vector, the torque distribution coefficient, mode switching threshold and control gain parameter are calculated in real time, and the first motor and the second motor are controlled to perform cooperative driving and mode switching.

[0017] Preferably, the constructed working condition feature vector includes:

[0018] Vehicle operation data is collected through the vehicle's CAN bus, on-board sensors, and motor controller of the pure electric bus.

[0019] The operating data includes at least vehicle speed, longitudinal acceleration, road gradient, battery state of charge, real-time torque of the first motor, real-time torque of the second motor, real-time speed of the first motor, and real-time speed of the second motor.

[0020] The operating data is collected synchronously at a fixed sampling frequency and time-aligned to form vehicle operating status data corresponding to the same moment.

[0021] Based on the vehicle operating status data, a condition feature vector is constructed to characterize the current driving state of the pure electric bus.

[0022] Preferably, the step of constructing the mapping relationship between operating condition characteristics and cooperative control parameters includes:

[0023] The collected historical operating data of pure electric buses were processed by time alignment, outlier removal and normalization.

[0024] Using the operating condition feature vector as input, the K-means clustering algorithm is used to perform cluster analysis on historical operating data, and the number of operating condition clusters is determined by the elbow rule.

[0025] Multiple operating condition clusters corresponding to typical operating states of pure electric buses are obtained. Each operating condition cluster corresponds to starting operating condition, acceleration operating condition, cruising operating condition, climbing operating condition or frequently switching operating condition.

[0026] For each operating condition cluster, its central eigenvector and covariance matrix are calculated to characterize the distribution characteristics of the vehicle's operating state under that operating condition.

[0027] Based on the distance relationship between the current working condition feature vector and each working condition cluster, the degree of membership of the current working condition to each working condition cluster is calculated.

[0028] Preferably, the construction of the dual-motor cooperative control parameter optimization model includes:

[0029] Define a dual-motor cooperative control parameter vector, which includes torque distribution coefficient, mode switching threshold, proportional control gain, integral control gain, derivative control gain, speed switching threshold, synchronization time window, and torque compensation coefficient.

[0030] Based on the current operating condition characteristic vector of the pure electric bus, combined with the efficiency spectrum, output torque range and battery power boundary of the first and second motors, the value range of each control parameter is determined respectively.

[0031] The value range is a dynamic upper and lower bound that changes with the vehicle's operating state, used to limit the feasible value range of the cooperative control parameters under the physical constraints of the dual-motor drive system.

[0032] Preferably, the multi-objective optimization includes:

[0033] The optimization targets are mode switching response time, mode switching impact, overall efficiency of dual-motor system, and motor life loss.

[0034] Under the premise of satisfying the motor output torque limit, battery power limit and system stability constraints, the cooperative control parameter vector is optimized and solved;

[0035] The optimization objectives are used to characterize the power performance, smoothness, energy efficiency, and reliability of pure electric buses during dual-motor drive and mode switching processes.

[0036] Preferably, the optimization model employs the Grey Wolf optimization algorithm.

[0037] During the population initialization phase, an initial population is generated based on the distribution characteristics of the cooperative control parameters under different operating conditions.

[0038] During the iteration process, the search for cooperative control parameters is achieved by updating the individual positions of gray wolves;

[0039] During the search process, the individual components that do not meet the constraints are corrected by combining the motor efficiency graph and torque constraints.

[0040] The optimal cooperative control parameter vector that meets the requirements of multi-objective optimization is obtained through iterative updates.

[0041] Preferably, the construction and use of the parameter-condition matching library includes,

[0042] The optimal collaborative control parameter vectors corresponding to each operating condition cluster are stored to form a parameter-operating condition mapping relationship;

[0043] During the operation of a pure electric bus, its membership degree to each operating condition cluster is calculated based on the current operating condition feature vector.

[0044] Select the cooperative control parameter vector corresponding to the operating condition cluster with the highest degree of membership as the current control parameter;

[0045] The control parameters are input into the vehicle controller to control the torque output and operating mode switching of the first and second motors.

[0046] A dual-motor coordinated control system for an electric bus, the system comprising:

[0047] The data acquisition module is used to acquire vehicle operating status data;

[0048] The operating condition identification module is used to construct operating condition feature vectors and identify the current operating condition;

[0049] The parameter optimization module is used to build a collaborative control parameter optimization model based on historical data and generate optimal parameters;

[0050] The parameter matching module is used to build a parameter-condition matching library and retrieve the optimal coordinated control parameters;

[0051] The control execution module is used to control the output torque and operating mode of the two motors according to the optimal cooperative control parameters.

[0052] An electronic device, including a processor and a memory,

[0053] The memory stores a computer program, and the processor executes the computer program to implement a method for coordinated control of dual motors in an electric bus.

[0054] A computer-readable storage medium,

[0055] The computer-readable storage medium stores a computer program that, when executed by a processor, implements a method for coordinated control of dual motors in an electric bus.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] (1) A full-process optimization mechanism based on working condition characteristics was constructed, and working condition information was introduced into the optimization and application process of vehicle control parameters, so that the control parameters can be dynamically adjusted according to the different operating states of pure electric buses such as starting, acceleration, cruising and climbing, thereby improving the response capability of the dual motor drive system to changes in the power demand of the vehicle and enhancing the adaptability of the vehicle control system under complex working conditions.

[0058] (2) In the optimization process, the present invention introduces the motor efficiency spectrum as guiding information, so that the dual motors tend to work in the high-efficiency range during the output torque distribution and mode switching process, effectively reducing power consumption, improving the overall vehicle operating efficiency, and avoiding the motor from being in a low-efficiency or high-load operating state for a long time, thereby improving the service life of the motor and power system.

[0059] (3) By classifying the constraints in the process of optimizing control parameters and introducing a robust evaluation mechanism in the parameter selection stage, the final coordinated control parameters not only meet the vehicle control constraints such as motor output capability, battery power and system stability, but also resist the influence of operating condition fluctuations and execution deviations in actual vehicle operation, thereby improving the stability and reliability of the vehicle control system.

[0060] (4) This invention constructs a parameter-condition matching mechanism for the whole vehicle controller of pure electric buses. By combining offline multi-condition optimization with online condition identification, the control parameters adapted to the current condition are called in real time during vehicle operation to realize the adaptive adjustment of dual motor drive mode and torque distribution strategy, ensuring the continuity and smoothness of mode switching process, reducing the impact of torque fluctuation, and improving ride comfort. Attached Figure Description

[0061] Figure 1 This is a flowchart of the method of the present invention;

[0062] Figure 2 This is a diagram showing the clustering results of the working conditions in this invention.

[0063] Figure 3 This is a comparison diagram of torque response during the mode switching process of the present invention;

[0064] Figure 4This is the adaptive weight allocation diagram for the working conditions of this invention; Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] This invention provides a technical solution:

[0067] S1. Problem Construction for Operational Data Acquisition, Operating Condition Feature Labeling, and Control Optimization

[0068] Operational data collection and condition feature annotation are carried out. The data collection process relies on the actual operation platform of the pure electric bus or a high-precision vehicle dynamics simulation model. The object of collection is the real-time operation data of the vehicle under various typical operating conditions.

[0069] The collected data specifically includes: vehicle speed, longitudinal acceleration, and road gradient, representing the vehicle's motion state; real-time torque of the first motor, real-time torque of the second motor, real-time speed of the first motor, real-time speed of the second motor, drive demand torque, and battery state of charge, representing the powertrain system state; mode switching flags and controller output parameters, representing the control execution process; in addition, to evaluate the control effect, motor efficiency data also needs to be collected synchronously for subsequent efficiency profile construction and analysis. The above data is collected synchronously via the vehicle's CAN bus, on-board sensors, and motor controller at a fixed sampling frequency to ensure time alignment and provide complete operational condition profile information for subsequent processing.

[0070] After data collection is completed, the data annotation stage begins. The purpose of annotation is to assign a clear operating condition category label to each set of operational data to support cluster analysis and operating condition identification. The annotation categories specifically include typical operating states such as start-up operating conditions, acceleration operating conditions, cruise operating conditions, hill climbing operating conditions, and frequent switching operating conditions.

[0071] The annotation method combines semi-automatic annotation with expert verification: First, based on the numerical range of key features such as vehicle speed, acceleration, and gradient, preliminary labels are automatically generated using rule-based algorithms. For example, segments where the vehicle speed increases from zero and the acceleration is greater than a set threshold are labeled as starting conditions, while segments where the vehicle speed remains stable and the absolute value of the acceleration is small are labeled as cruising conditions. Then, calibration engineers sample, review, and correct the automatic annotation results to ensure the accuracy of condition classification. In particular, for transitional conditions with frequent mode switching and blurred feature boundaries, manual intervention is used to ensure the consistency between the labels and the actual operating state of the vehicle, thereby forming a high-quality historical operating dataset with accurate condition labels.

[0072] Based on this, for dual-motor pure electric buses during rapid driving and mode switching, an optimal set of cooperative control parameter vectors is sought to comprehensively optimize four key performance indicators—mode switching response time, switching process impact, overall efficiency of the dual-motor system, and motor life loss—while meeting the hard requirements of motor physical constraints, battery power boundaries, and system stability. Since the preferences for these four indicators differ significantly under different operating conditions, and the control parameters exhibit a strong nonlinear coupling relationship with operating condition characteristics, this optimization problem is essentially a complex parameter optimization problem oriented towards multiple objectives, strong constraints, and time-varying operating conditions.

[0073] S2. Hierarchical encoding and initial population construction of dual-motor cooperative control parameters based on operating condition characteristics.

[0074] In actual operation of pure electric buses, the torque distribution coefficient, mode switching threshold, proportional-integral-derivative control gain, synchronization time window, and torque compensation coefficient of the dual-motor drive system exhibit significantly different sensitivities depending on changes in vehicle speed, acceleration, road gradient, battery state of charge, and the real-time torque and speed of the dual motors. Conventional gray wolf optimization algorithms typically use uniform random sampling within fixed upper and lower bounds when initializing the population. This approach fails to reflect the differences in parameter distribution under different operating conditions and does not utilize engineering constraints such as motor efficiency maps, torque upper limits, battery power boundaries, and switching smoothness requirements. This easily leads to a large number of initial individuals falling into infeasible or low-quality regions, resulting in low subsequent optimization efficiency and unstable convergence.

[0075] First, establish the correspondence between the collaborative control parameter vector and the operating condition feature vector. Then, identify the sensitive areas of the operating conditions based on historical operating data. Finally, construct an initial population that conforms to the operating condition rules using dynamic boundaries and probability distributions, so that the initial solution preferentially covers the key feasible areas under various operating conditions. The specific steps are as follows:

[0076] 1> Define the cooperative control parameter vector and working condition feature vector ,as follows:

[0077] ;

[0078] ;

[0079] in, The vector of cooperative control parameters to be optimized has a dimension of 8 and is used to characterize the combination of core control parameters of the dual-motor drive system during rapid driving and mode switching.

[0080] This represents the torque distribution coefficient, which ranges from 0 to 1 and is used to determine the target torque distribution ratio between the first motor and the second motor.

[0081] Indicates the mode switching threshold, in units of This is used to determine whether the switch between single-motor mode and dual-motor mode is triggered.

[0082] , and The proportional-integral-derivative (PI-DE) parameters that together constitute the torque coordination controller for the dual motors, wherein, Indicates proportional control gain. Indicates integral control gain. Indicates the differential control gain;

[0083] Indicates the speed switching threshold, in units of , used to define the speed boundaries of different motor operating areas;

[0084] Indicates the synchronization time window, in units of This is used to specify the time range within which the torques of the two motors reach consistency during constraint mode switching.

[0085] This represents the torque compensation coefficient, used for feedforward compensation during mode switching to suppress torque drop or torque surge caused by the switching.

[0086] This represents the current operating condition feature vector, with a dimension of 8, used to comprehensively characterize the current driving state of the pure electric bus.

[0087] Indicates vehicle speed, in units of ; Represents the longitudinal acceleration of a vehicle, in units of ;

[0088] Indicates road gradient, unit: ;

[0089] This indicates the battery's state of charge, with a value ranging from 0 to 1.

[0090] This indicates the real-time output torque of the first motor, in units of... ; This indicates the real-time output torque of the second motor, in units of... ;

[0091] This indicates the real-time speed of the first motor, in units of... ; This indicates the real-time speed of the second motor, in units of... .

[0092] 2> Construct sensitive operating conditions areas based on historical operational data.

[0093] First, historical data of the dual-motor pure electric bus under typical operating conditions is collected. This historical data includes at least vehicle speed, acceleration, gradient, battery state of charge, dual-motor torque, dual-motor speed, drive torque demand, mode switching flags, motor efficiency, and controller output parameters. Then, the historical data undergoes time alignment, outlier removal, and normalization. Finally, a feature vector based on the operating conditions is used. Using the data as input, cluster analysis is performed on the samples to obtain multiple operating condition clusters. Each operating condition cluster corresponds to a typical operating condition region, such as starting operating condition, acceleration operating condition, cruising operating condition, climbing operating condition, and frequently switching operating condition.

[0094] Then, for each cluster of operating conditions (indicated by the first...) Taking a cluster of working conditions as an example, calculate its central feature vector. and the degree of discreteness matrix ,in, Characterizing the first The mean vector of each working condition cluster represents the typical characteristics of that working condition cluster. Characterizing the first The covariance matrix of each operating condition cluster represents the fluctuation range and correlation of each characteristic quantity within that operating condition cluster. This indicates the operating condition cluster number.

[0095] When performing time alignment, outlier removal, and normalization on historical data, linear interpolation is used for time alignment, the 3σ principle (Laida criterion) is used for outlier removal, and minimum-maximum normalization is used for normalization.

[0096] K-means clustering algorithm is used for clustering, and elbow rule is used to determine working condition clusters. The center of each cluster is considered to be the typical feature vector of that working condition.

[0097] Based on the current operating condition feature vector The distance relationship with each working condition cluster is used to calculate the degree of membership of the current working condition to each working condition cluster. The higher the degree of membership, the closer the current working condition is to that working condition cluster, and the more the corresponding prior parameter distribution should be close to the historical optimal parameter region of that working condition cluster.

[0098] The distance relationship is preferably estimated using Gaussian kernel density because this method can reflect both the center position of the operating condition and the joint distribution of various operating condition variables, making it more suitable for describing the operating condition distribution characteristics of multivariate coupling in a dual-motor drive system. Furthermore, the membership degree is calculated using the Gaussian kernel function.

[0099] 3> Based on the working condition feature vector Determine the coordinated control parameter vector The dynamic upper and lower bounds of each parameter: This invention does not set fixed boundaries for the eight parameters, but rather generates allowable ranges for each parameter based on the current operating condition. Specifically,

[0100] for Based on the efficiency graphs of the first and second motors, a distribution region with relatively high overall efficiency is identified near the current speed and target torque. The upper and lower bounds are then determined by combining this with the maximum allowable torque of both motors. For example, assuming the current required torque is... Motor speed 1 Motor speed 2 On the efficiency graph, find the torque point where the two motors have the highest efficiency at their respective speeds. For example, the high-efficiency region of motor 1 is at... Nearby, the high-efficiency zone of motor 2 is in Nearby, the total efficient torque is ,So, The lower bound can be set to The upper bound is set to This forms a narrow interval that is biased towards maximizing overall efficiency;

[0101] for and The upper and lower bounds are determined based on mode switching frequency limits, power response requirements, and the distribution of the motor's high-efficiency zone. For example, in climbing conditions, to avoid power interruption caused by frequent switching, single-to-dual motor switching at low speeds and high torque should be limited. The upper bound is relatively high (e.g.) This means that only when the required torque exceeds Dual-motor mode is only allowed at certain times to maintain stable power. During high-speed cruising, to enter the high-efficiency range of a single motor, the following settings can be configured: The lower bound is lower (e.g.) This means that when the motor speed is higher than this value, the system will try to switch to single motor mode first.

[0102] for , and The upper and lower bounds are determined based on torque tracking bandwidth, system oscillation risk, and existing controller calibration experience. For example, referring to existing vehicle controller calibration data, for torque tracking control... The value range is usually in arrive In between, and under rapid acceleration conditions, in order to respond quickly, it can be The upper bound for allowance is expanded to This allows for a larger proportion of action, sacrificing some stability in exchange for faster response speed;

[0103] for The upper and lower bounds are determined based on the requirements for smoothness and response speed. For example, in a start-up condition where smoothness is prioritized, to ensure a smooth switching process, the upper and lower bounds can be set. The allowed upper bound is This allows for a longer synchronization time for a smooth transition, while in overtaking situations where power is prioritized, a faster mode switch can be implemented. The lower bound is allowed to tighten to Limiting the maximum synchronization time forces the controller to act quickly;

[0104] for The upper and lower bounds are determined based on the expected torque gap, inertia compensation requirements, and upper limit of impact at the moment of switching. For example, at the moment of switching, if it is predicted that the disengagement of motor 1 will cause... The torque gap, then The lower bound can be set to Ensure at least compensation To maintain stable total output torque, and to avoid excessive impact due to compensation overshoot, an upper limit for impact is considered. Given the limitations, the maximum allowable compensation coefficient can be calculated, let it be denoted as . The upper bound, for example .

[0105] In specific implementation, the first The dynamic lower bound of each parameter under the current operating conditions is denoted as... The dynamic upper bound is denoted as ,in, Characterizing the first Each parameter in the working condition feature vector The minimum allowed value, Characterizing the first Each parameter in the working condition feature vector The maximum allowed value; This indicates the parameter dimension number, with values ​​ranging from 1 to 8.

[0106] The dynamic lower and upper bounds are not simple fixed empirical values, but are jointly determined by the motor efficiency spectrum, the output torque range of the dual motors, the battery discharge capacity, the controller stability requirements, and the mode switching smoothness requirements.

[0107] An efficiency map is a two-dimensional contour map that describes the working efficiency of a motor under different speeds and different combinations of output torque. It intuitively shows the high-efficiency operating area (efficiency island) and low-efficiency operating area of ​​the motor, and is the core basis for motor selection, control and energy management optimization.

[0108] The analysis of the operating condition clustering results shows the distribution of vehicle operation data in the PCA dimensionality reduction space after clustering using the K-means algorithm. Different colors represent different operating condition clusters, and the red crosses are the cluster centers. Historical operation data can be divided into multiple typical operating condition categories (such as starting, acceleration, cruising, climbing, and frequent switching). The horizontal and vertical axes are Principal Component 1 and Principal Component 2 (dimensionless, feature dimensions after PCA projection), respectively. Different colors represent different operating condition clusters.

[0109] 4> Based on parameter sensitivity, the cooperative control parameter vector Hierarchical coding is performed, and an initial population is generated based on the prior distribution of working conditions. Specifically,

[0110] First, divide the eight parameters into three groups based on function and sensitivity, including:

[0111] The first group consists of energy allocation and switching boundary parameters, including , and These parameters mainly affect overall efficiency and the timing of switching triggers;

[0112] The second group consists of the controller's dynamic parameters, including... , and These parameters mainly affect torque tracking speed, overshoot, and oscillation.

[0113] The third group is for switching coordination parameters, including and These parameters mainly affect the synchronization speed and torque smoothness during the switching process.

[0114] After grouping and encoding, different groups of parameters can use different sampling biases, making the initialization process closer to the real and effective parameter distribution under various working conditions.

[0115] For each individual gray wolf and each parameter dimension, the sampling bias is first determined based on the degree of membership in the current working condition cluster. Then, random initial values ​​are generated within the dynamic boundary of the current parameter. To balance boundary coverage and sensitive area clustering, random numbers are preferentially generated using a Beta distribution. Specifically...

[0116] set up Indicates the first The first gray wolf individual in the... Random numbers distributed across each parameter dimension, with values ​​ranging from 0 to 1; This represents the individual number of the gray wolf; then the... The first gray wolf individual in the... The initial values ​​for each parameter dimension can be represented as: ;

[0117] in, Indicates the first in the initial population The first gray wolf individual in the... Initial values ​​for each parameter dimension; Indicates the relationship with the first The working condition feature vector corresponding to each individual gray wolf; This represents the interval proportion value obtained by sampling from the Beta distribution; The term is used to scale the interval width so that the initial parameter values ​​strictly fall within the dynamic boundaries.

[0118] The two shape parameters of the Beta distribution are denoted as follows: and ,in, Indicates the shape parameter on the left side of the distribution. The shape parameter on the right side of the distribution can be used to establish a mapping table from operating condition type to the shape parameter of the Beta distribution, thereby adjusting the probability density. Specifically, for start-up and rapid acceleration conditions, the distribution can be shifted towards the high-response region. , and smaller It is easier to sample; for cruise conditions, it can shift the distribution towards high-efficiency regions, making... and It is easier to fall into the region with higher overall efficiency; for frequently switching operating conditions, it can cause the distribution to shift towards the low-impact region, making... and It is easier to land in areas with better smoothness.

[0119] 5. Perform feasibility checks and adjustments on the initial population to obtain an initial gray wolf population suitable for iteration. Specifically,

[0120] Perform rapid simulation or approximate evaluation on each individual gray wolf to check for violations of hard constraints. Hard constraints include at least the maximum torque limit of the first motor, the maximum torque limit of the second motor, the maximum discharge power limit of the battery, and the maximum speed limit of the motor.

[0121] When an individual violates a hard constraint, the projection correction method is used first to pull the violation parameters back to the nearest feasible boundary; if the constraint still cannot be satisfied after correction, the individual is resampled.

[0122] For soft constraints, such as excessive mode switching impact, excessive torque fluctuation, or excessively long switching time, the individual can be temporarily retained, but its soft constraint risk level should be recorded to provide a basis for subsequent penalty calculation.

[0123] The high-quality parameters of a dual-motor drive system are not uniformly distributed throughout the parameter space, but are concentrated in local areas that meet the requirements of efficiency, power, and smooth switching. This invention does not simply generate an initial population randomly in a fixed parameter space, but first determines the parameter-sensitive area according to the working conditions, then performs hierarchical encoding according to the parameter functions, and finally initializes it by combining dynamic boundaries and prior distributions of working conditions. This method can significantly increase the probability of the initial population falling into the critical feasible region, thereby improving the efficiency of subsequent optimization and reducing the time consumption caused by a large number of invalid iterations in the early stage.

[0124] S3. Improved working condition adaptive search mechanism and cooperative constraint handling of the gray wolf optimization algorithm

[0125] The problem of optimizing the cooperative control parameters of a dual-motor switching pure electric bus is characterized by strong nonlinearity, multi-constraint coupling, and time-varying operating conditions. Traditional gray wolf optimization algorithms typically use linearly decreasing convergence factors and fixed leader wolf weights, which cannot dynamically adjust the exploration and development intensity according to the complexity of the current operating conditions. They also fail to incorporate motor efficiency maps, torque response characteristics, and physical constraints of mode switching into the position update process, making them prone to getting stuck in local optima or generating infeasible solutions in the later stages of iteration.

[0126] The Grey Wolf optimization framework introduces a working condition-adaptive nonlinear convergence factor, an efficiency guiding term, and a hierarchical penalty mechanism to enable the search process to maintain global exploration capabilities while rapidly converging to a high-quality parameter region under engineering constraints. The specific steps are as follows:

[0127] 1> At the beginning of each iteration, based on the current working condition feature vector Calculate the adaptive nonlinear convergence factor under the operating condition. Specifically, let... Indicates the current iteration number. Indicates the maximum number of iterations. Indicates the first The convergence factor at the next iteration This represents the initial convergence factor, which is typically set to 2.

[0128] First, a condition complexity index is constructed based on acceleration, gradient, and mode switching frequency. Higher condition complexity indicates more local optima in the current search space, necessitating a longer global search phase to slow the convergence factor decline. Lower condition complexity indicates a smoother search space, requiring faster local convergence. Then, a small periodic perturbation is superimposed on the basic decay curve to prevent premature rigidity of the population in the later stages of iteration, which can be expressed as: ;in, The adaptive decay exponent for the operating condition is used to control the convergence rate. It is obtained by normalizing the acceleration and gradient. For example, a smaller value is taken under high acceleration and steep gradient conditions, and a larger value is taken under cruise conditions, thereby achieving the effect of slow convergence under complex conditions and fast convergence under simple conditions. This represents the disturbance amplitude, which can be taken as 0.02 to 0.08. Indicates the period of disturbance; and These are the normalized values ​​of the maximum acceleration and maximum gradient that may occur under the operating conditions.

[0129] 2> Based on the fitness results after adding constraint penalties, determine Wolf, wolves and Wolves, and calculate the dynamic weights of the three alpha gray wolves, specifically,

[0130] First, for each individual gray wolf in the current population, the vehicle-motor-controller co-simulation model or the fast proxy model is called one by one to obtain its comprehensive fitness value and constraint violation degree under the current working conditions. Then, the comprehensive fitness value and the constraint penalty term are combined into a penalty fitness value (synthesized by addition), and the individuals are sorted from best to worst according to their penalty fitness values. The individual with the best penalty fitness value is denoted as... The second-best individual is denoted as The third best individual is denoted as ,Right now, , and They represent the first The collaborative control parameter vector of the three leader wolves in the next iteration;

[0131] Then, when calculating the dynamic weights of the three leader wolves, the weights of the three leader wolves are not fixed values, but dynamic weights related to the penalty fitness value. In order to ensure that individuals with better fitness have greater weights, the weights are constructed using the reciprocal of the penalty fitness value. That is, the smaller the penalty fitness value, the greater the corresponding weight. This can avoid the problem of the contribution ratio of the three leader individuals being too rigid in the original gray wolf optimization algorithm, and make the iteration direction closer to the current true high-quality solution region.

[0132] In the gray wolf optimization algorithm, α wolf, β wolf, and δ wolf are the three individuals with the best fitness in the population. They represent the three best solutions found so far, respectively simulating the alpha wolf, the co-alpha wolf, and the leadership of the ordinary wolf pack. Other gray wolf individuals (ω wolf) will refer to the positions of these three leaders when updating their own positions, thereby achieving convergence towards the optimal solution region.

[0133] The vehicle-motor-controller co-simulation model or fast proxy model refers to an evaluator used to assess the performance of each individual wolf (a set of parameters). It assumes the existence of a high-fidelity simulation model that can simulate the dynamic response of a dual-motor pure electric bus based on the input cooperative control parameters Θ and operating condition characteristics C, and output the various performance indicators (such as response time, impact, etc.) and constraint violation degree required by the fitness function. In practical applications, this model is usually jointly built by vehicle dynamics software (such as CarSim, AMESim) and motor control model (such as Simulink).

[0134] The overall fitness value and constraint violation degree under the current operating condition can be obtained by calling the aforementioned vehicle-motor-controller joint simulation model. Specifically, the cooperative control parameter vector Θ represented by the individual gray wolf and the current operating condition feature vector C are used as inputs to run the simulation model. After the model simulation is completed, predefined indicators, such as response time, are extracted from the output results. Impact And so on, and calculate the degree of constraint violation based on preset constraints (such as maximum torque limit).

[0135] 3> In each iteration, the position of each individual wolf is updated based on the information of the leader wolf, the current global optimal solution, and the motor efficiency graph guidance information. Specifically,

[0136] First use , and The dynamic weighted average result is used to construct the basic update direction (based on the dynamic weights of the three leader wolves), causing the group as a whole to move closer to the current high-quality area; then, an attraction term for the global optimal position is introduced, causing individuals to gradually shrink to a better area; then, a heavy-tailed random perturbation is superimposed, allowing some individuals to still make long-distance jumps, thereby improving their ability to jump out of the local optimum; finally, an efficiency guidance term is superimposed, causing the update direction to preferentially move towards the parameter area with higher overall efficiency of the dual motors.

[0137] The location update method for individual gray wolves is represented as follows: ;

[0138] in, ;

[0139] ;

[0140] ; ;

[0141] Indicates the first The first gray wolf individual in the... The position at the next iteration (i.e., the cooperative control parameter vector); Indicates the first The positions of α, β, and δ wolves at the next iteration; This represents the dynamic weights of the three alpha wolves, which are inversely proportional to their fitness values; Indicates as of the date The globally optimal position found in the next iteration; Indicates the first The first gray wolf individual in the... The current position at the next iteration; This represents the adaptive step size factor, used to control the step size for moving towards the global optimum. It typically decreases with the number of iterations, such as... . This represents a heavy-tailed random perturbation vector, where each dimension is independently sampled from a standard Cauchy distribution to produce larger jumps and help escape local optima. The attenuation coefficient representing the disturbance decreases as the number of iterations increases, in order to balance exploration and exploitation. ; This represents the learning rate of the efficiency-guided term, controlling the strength of the guidance; a preferred value is 0.01. This represents the efficiency-guided vector, whose direction points to the parameter changes that improve the overall efficiency of the dual-motor system. Specifically, it is determined by... , and Small positive and negative disturbances are applied to key parameters, and the changes in the overall system efficiency before and after the disturbances are calculated by interpolation using the motor efficiency spectrum. Then, the gradient direction of the parameter changes with respect to efficiency is estimated, thereby constructing the efficiency guiding vector.

[0142] The overall system efficiency can be calculated from the relationship between the mechanical output power of the first motor, the mechanical output power of the second motor, and the drive demand power, where,

[0143] The mechanical output power of the first motor is determined by and The mechanical output power of the second motor is determined jointly (through product calculation) by... and The efficiency of the first motor and the efficiency of the second motor are determined together (by multiplication); the efficiency of the first motor and the efficiency of the second motor are obtained by interpolation from their respective efficiency graphs.

[0144] The rotational speed of motor 1 at a certain moment is The output torque is Its efficiency graph is a two-dimensional table with speed and torque as coordinates. First, four adjacent speed points (e.g., 900 and 1000 rad / s) and four adjacent torque points (e.g., 200 and 220 N·m) are found in the graph to define a grid. Then, using bilinear interpolation, the corresponding efficiency value is calculated based on the relative positions of the current speed and torque in the grid. .

[0145] The role of the efficiency guide term is not to replace the gray wolf optimization algorithm, but to add a directional correction that conforms to the physical mechanism of dual motors on the basis of the original group cooperative search, so that the algorithm can not only "move towards a better fitness", but also "move towards a more reasonable dual motor efficiency region".

[0146] 4> Perform hierarchical constraint processing based on dual-motor switching physical constraints on the updated gray wolf individuals. Specifically,

[0147] Constraints are divided into two categories: hard constraints and soft constraints.

[0148] Hard constraints include at least the maximum torque limit of the first motor, the maximum torque limit of the second motor, the maximum discharge power limit of the battery, and the maximum speed limit of the motor. Violation of hard constraints may lead to system protection action, device overload, or safety risks. Therefore, they must be prioritized to be met.

[0149] Soft constraints include at least mode switching impact constraints, torque fluctuation constraints, motor temperature rise constraints, and synchronization time constraints. Soft constraints are tolerable within a certain range, but the greater the degree of violation, the worse the control quality.

[0150] First, perform a quick feasibility check on the updated parameter vector. If it violates hard constraints, prioritize direct correction methods, such as reducing the size of the parameter vector. Variable range, improvement To reduce low-speed, high-torque switching and increase To reduce the synchronization rate, or to directly project the corresponding parameters to the nearest feasible boundary, if an individual still cannot satisfy the hard constraints after correction, it is considered an invalid individual in this round and is regenerated. For soft constraints, individuals are not directly deleted, but a penalty term is added during fitness evaluation. The penalty coefficient for hard constraints gradually increases with iteration, making the algorithm converge more strictly to the physical feasible region in later stages. The penalty coefficient for soft constraints is moderate in the early stage of iteration and can maintain a smooth change in the later stage, giving the algorithm the opportunity to explore better solutions near the feasible boundary.

[0151] The purpose of hierarchical constraint processing is to ensure that the optimized parameters are not only mathematically optimal but also physically feasible. Its core is to introduce a penalty fitness value, which transforms the degree of constraint violation into a penalty for the fitness value, guiding the search towards the feasible region. The penalized fitness value is defined as follows: ;

[0152] in, ;

[0153] ;

[0154] This represents the original, unconstrained overall fitness value. Represents a set of hard constraints; Represents the set of soft constraints; Indicates the first A function for the degree of violation of a hard constraint, for example, for a maximum torque constraint. ; Indicates the first A penalty function for each soft constraint is used to calculate the degree of violation, for example... ; Indicates the first The penalty coefficient for each hard constraint varies with the number of iterations. Increase and grow exponentially (e.g.) This forces the algorithm to strictly satisfy hard constraints in its later stages; Indicates the first The penalty coefficient for each soft constraint is a fixed value, balancing the exploration of feasible solutions with the optimization of the objective function.

[0155] In practical implementation, to avoid abrupt changes in the penalty term near the constraint boundary, a smooth penalty function is preferred to handle soft constraint violations. Specifically, a smooth penalty is implemented through a quadratic penalty function. That is, when the violation is small, the penalty increases slowly, which helps the algorithm retain high-quality solutions near the boundary. When the violation is large, the penalty is significantly enhanced, thereby suppressing obviously unreasonable solutions.

[0156] 5. Repeat the fitness evaluation, leader selection, position update, and constraint handling until the termination condition is met. The termination condition may be reaching the maximum number of iterations. Alternatively, it could be that the change in the global optimal solution is lower than a preset threshold during multiple consecutive iterations.

[0157] Meanwhile, in order to provide candidate solutions for subsequent multi-objective decision-making, a non-dominated solution archive is maintained simultaneously during the iteration process. For any two individuals, if one individual is no worse than the other individual in all four indicators of response time, impact, overall efficiency and motor life loss, and is better in at least one indicator, then the former is considered to dominate the latter. All solutions not dominated by other individuals together constitute the Pareto optimal candidate set.

[0158] The optimization of collaborative control parameters for dual-motor switching pure electric buses is not only a mathematical search problem, but also an engineering problem constrained by efficiency, power, torque and ride comfort. This invention integrates adaptive convergence control, efficiency map guidance and hierarchical processing of physical constraints into the Grey Wolf optimization algorithm. This avoids the algorithm blindly searching in the abstract parameter space, but instead ensures efficient evolution in a direction that conforms to the dual-motor drive mechanism, thereby improving the convergence speed and the engineering usability of the solution.

[0159] A comparison of torque response during mode switching was conducted. By simulating a dynamic process of switching from single-motor mode to dual-motor mode, the advantages of this technology over conventional technologies in torque response speed and tracking accuracy were demonstrated. The experiment was set up using a vehicle dynamics co-simulation model, with the input target torque curve jumping from 200 Nm to 500 Nm in 0.2 seconds to simulate the driver's rapid acceleration needs. Figure 3 The horizontal axis represents time (seconds), and the vertical axis represents output torque (N·m). The three curves represent the target torque, the output torque of this technology, and the output torque of conventional technology, respectively. The target torque is represented by a black dashed line, the output torque of this technology by a blue solid line, and the output torque of conventional technology by a red solid line. Conventional technology refers to a control method using a fixed proportional-integral-derivative gain and a fixed synchronization time window, without introducing an adaptive nonlinear convergence factor or efficiency guide term. Its parameters are calibrated based on typical operating conditions, but it lacks the ability to actively suppress torque fluctuations and oscillations during transient switching. This technology, on the other hand, calculates the efficiency guide vector and adaptive step size factor in real time, enabling the torque controller to quickly track the target value at the moment of switching, while using dynamic adjustment of the torque compensation coefficient and synchronization time window to suppress overshoot and oscillations. Judging from the curve shape, the output torque of this technology (blue line) enters the 5% error band of the target torque within about 0.05 seconds after the step occurs, and only slight fluctuations occur throughout the process before stabilizing rapidly. In contrast, the conventional technology (red line) has a significantly longer rise time. This indicates that the cooperative control parameters obtained by this invention through the improved Grey Wolf optimization algorithm can achieve faster torque response, smaller overshoot, and shorter settling time during mode switching, thereby simultaneously meeting the dual requirements of rapid drive and smooth switching.

[0160] S4. Construction of a comprehensive fitness function and optimal decision-making for multi-objective cooperation in dual-motor switching.

[0161] The coordinated control of a dual-motor drive system during mode switching involves multiple objectives, including response speed, smoothness, energy efficiency, and actuator durability. Furthermore, different operating conditions exhibit significantly different preferences for these objectives. Urban congestion conditions prioritize impact and smoothness, high-speed overtaking conditions prioritize response speed, long-distance cruising conditions prioritize overall efficiency, and frequent start-stop conditions prioritize motor lifespan and thermal load. Conventional fixed-weight single-objective optimization methods are ill-suited to reflect these differences in operating conditions.

[0162] A multi-objective comprehensive fitness function driven by working condition identification is constructed. Dynamic weighting is implemented in the search phase, and a more robust final parameter solution is selected from the Pareto optimal candidate set in the decision phase, thereby taking into account the engineering requirements under different working conditions. The specific steps are as follows:

[0163] 1> Calculate the response time index, impact index, overall efficiency index, and motor life loss index separately, and normalize the four indices to the same dimension. Specifically,

[0164] Response time is marked as This indicates the time elapsed from receiving the mode switching command until the total output torque of the two motors first stabilizes at around 95% of the target torque;

[0165] Impact intensity is marked as , which represents the peak value of the rate of change of the vehicle's longitudinal acceleration during mode switching, can be obtained by differentiating the vehicle's longitudinal acceleration curve and taking the maximum absolute value;

[0166] Overall efficiency is marked as This represents the equivalent energy consumption level of the dual-motor drive system within the mode switching window. To unify the minimization direction, the overall system efficiency can be... Convert to ,in, The overall efficiency of a dual-motor drive system is the ratio of the total mechanical power output of the two motors to the total electrical power obtained from the battery during mode switching.

[0167] Motor life loss is indicated by This indicates the degree of equivalent life damage caused by torque fluctuations, speed fluctuations, and temperature rise during the switching process, which can be calculated based on the motor load time history.

[0168] Motor life loss index The calculation is based on the motor load time history. Specifically, the torque curve and speed curve of the dual motors in the mode switching window are extracted first. Then, the equivalent load cycle number is identified using the rain flow counting method. Next, the temperature rise level is obtained by combining the motor thermal model. Finally, the load cycle intensity and temperature rise level are mapped to the life loss value. It should be noted that a specific life model is not required here. As long as it can stably represent the relationship that "the more intense the switching, the higher the thermal load, and the greater the life loss", it is acceptable.

[0169] To eliminate the influence of different dimensions among the four indicators, normalization is performed on them. Normalization can use the current population minimum and maximum values, historical calibration extreme values, or typical operating condition reference extreme values ​​as the upper and lower bounds. After normalization, the following results are obtained: , , and All are dimensionless indices between 0 and 1, with smaller values ​​indicating better performance. This represents the normalized response time metric; the smaller the value, the faster the mode switching response. This represents the normalized impact index. The smaller the value, the better the transition smoothness and the higher the ride comfort. This represents the normalized overall efficiency index; the smaller the value, the higher the energy efficiency of the drive system. This represents the normalized motor life loss index. The smaller the value, the less damage the switching process causes to the motor's life.

[0170] 2> Based on the current operating condition feature vector Calculate the adaptive weights for the four performance indicators under specific operating conditions.

[0171] Multiple operating condition templates are pre-established, each corresponding to a typical operating state, such as starting, acceleration, cruising, braking, and climbing. The feature vector of each working condition template is denoted as . This represents the representative characteristics of the typical operating condition in terms of vehicle speed, acceleration, gradient, battery state of charge, and dual-motor operation; the benchmark weight vector corresponding to this operating condition template is denoted as... This is used to indicate the importance preference of four indicators under this working condition, among which, Indicates the first The baseline weight of the response time indicator in each working condition template. Indicates the first The baseline weight of the impact index in each working condition template. Indicates the first The benchmark weight of the comprehensive efficiency index in each working condition template. Indicates the first The benchmark weight of motor life loss index in each working condition template.

[0172] In the specific implementation, the feature vector of the current working condition is first calculated. With each working condition template feature vector The similarity between templates is calculated; the higher the similarity, the closer the current working condition is to the template. Similarity can be obtained using a Gaussian kernel function, the reciprocal of the normalized Euclidean distance, or other continuous similarity functions. Then, using the similarity of each template as a fusion coefficient, a weighted average is calculated on the baseline weight vector of each template, followed by normalization, to obtain the dynamic weights under the current working condition. , , and ,in, This indicates the weight of the response time metric under the current operating conditions. This indicates the weight of the impact index under the current operating conditions; This indicates the weight of the overall efficiency index under the current operating conditions. This represents the weight of the motor life loss index under the current operating conditions, and the sum of the four weights is 1.

[0173] Assume there are only two operating condition templates: starting condition (i=1) and cruise condition (i=2), and their baseline weights are as follows:

[0174] Starter template: (Impact is the most important factor);

[0175] Cruise template: (Efficiency is the most important factor);

[0176] The current operating condition C, after calculation, has a similarity of [value] to the starting template. The similarity with the cruise template is The fusion weights are then:

[0177] ;

[0178] ;

[0179] ;

[0180] ;

[0181] Finally, the four weights are normalized so that their sum is 1, thus obtaining the dynamic weights for the current working condition.

[0182] The baseline weight vectors for each working condition template can be predetermined using the fuzzy hierarchical analysis method. Specifically, the calibration personnel first compare the importance of response time, impact, overall efficiency and motor life loss pairwise, then use triangular fuzzy numbers to describe this comparison relationship, and finally obtain the template weights through consistency verification and defuzzification. This allows engineering experience and working condition differences to be reflected in the weight generation process, avoiding the problem of insufficient adaptability to real working conditions when using fixed weights.

[0183] This invention demonstrates its ability to dynamically allocate weights to four performance indicators based on real-time operating conditions, proving that the comprehensive fitness function is no longer a fixed weighted sum, but can automatically adjust preferences as operating conditions change. The horizontal axis represents four typical operating conditions, the vertical axis represents the sum of the weights of each indicator, and the total height is always 1. Figure 4 Each column is composed of four differently colored blocks, representing the weights for response time, impact, overall efficiency, and motor lifespan loss, respectively. The dynamic weight calculation method based on the similarity fusion of operating condition templates in this invention can accurately reflect engineering preferences under different operating conditions, avoiding the shortcomings of traditional methods where fixed weights cannot accommodate various scenarios.

[0184] 3> Construct a comprehensive fitness function based on normalized indices and adaptive weights for operating conditions, and use it as the core evaluation criterion for the search phase. The comprehensive fitness function is expressed as follows: ,in, Represents the coordinated control parameter vector In the working condition feature vector The overall fitness value is given below; the smaller the value, the better the overall performance.

[0185] Then, in S3, the integrated fitness function is combined with the hierarchical constraint penalty to form a penalty fitness value for iterative sorting. This can reflect the preference of different working conditions for different objectives in the search phase, and can also avoid the physical infeasibility solution being mistakenly selected because a certain single index is better.

[0186] It should be noted that after initializing the population and before starting the iteration, an individual evaluation can be performed first, which is used for sorting gray wolves and selecting the leader wolf in the first iteration.

[0187] 4> After the iteration is complete, select the final optimal cooperative control parameter vector from the Pareto optimal candidate set. This invention does not arbitrarily select a solution from the Pareto optimal candidate set, but further performs an improved sorting process to approximate the ideal solution. Specifically,

[0188] First, the dispersion of the four indicators in the Pareto optimal candidate set is statistically analyzed, and the objective weights of the four indicators are calculated using the entropy weight method. The greater the difference of an indicator in the Pareto optimal candidate set, the stronger its contribution to distinguishing the quality of candidate solutions, and the higher its corresponding objective weight. Then, the ideal value and anti-ideal value of the four indicators are determined respectively. The ideal value represents the best level of the indicator in the Pareto optimal candidate set, and the anti-ideal value represents the worst level of the indicator in the Pareto optimal candidate set.

[0189] Then, for each candidate solution in the Pareto optimal candidate set, calculate its distance to the ideal solution and its distance to the anti-ideal solution. When calculating the distance, normalized weighted distance is preferred to avoid the excessive influence of a certain indicator dimension or fluctuation range on the result. The closer the distance is to the ideal solution and the farther it is from the anti-ideal solution, the better the comprehensive performance of the candidate solution.

[0190] Then, based on the ranking of approximate ideal solutions, a robustness evaluation is introduced to obtain the final optimal cooperative control parameter vector. Specifically, small perturbations are applied to each parameter of each candidate solution, for example, to , , , , , , and By adding a small positive and negative perturbation respectively, and recalculating the changes in the four performance indicators, if the changes in the four performance indicators are still small after a small shift in the parameters of a candidate solution, it indicates that the candidate solution is not sensitive to manufacturing errors, execution errors and fluctuations in actual operating conditions, and has stronger robustness. Conversely, if a candidate solution is extremely sensitive to parameter perturbations, even if the static evaluation is better, it is not suitable for direct use in engineering applications.

[0191] In practical implementation, the performance distance evaluation results and robustness evaluation results can be weighted and fused to obtain the final decision score for each candidate solution; the candidate solution with the best decision score is denoted as... , The final selected optimal cooperative control parameter vector is used as the parameter configuration result of the cooperative controller for the dual-motor drive system; output Afterwards, it can be directly written into the controller parameter table, or it can be combined with the operating condition template to form a parameter lookup table library, which can be quickly called up according to the current operating condition when the vehicle is running online.

[0192] In practical implementation, the performance distance evaluation result refers to the distance from each candidate solution to the ideal solution and the anti-ideal solution. It is obtained by calculating the weighted Euclidean distance of each candidate solution in the process of approximating the ideal solution and sorting it. Specifically, for each solution in the Pareto optimal candidate set, two distances are calculated: (Distance to the ideal solution) and (The distance to the antiideal solution) These two distances together constitute the performance distance evaluation result of the solution.

[0193] In practical implementation, the robustness evaluation result refers to the sum (or weighted sum) of the changes in the four performance indicators of each candidate solution after the parameters are subjected to a small range of perturbation. In the robustness evaluation step, the four indicators are recalculated by applying a small perturbation (such as ±5%) to the parameters of each candidate solution, and the results are compared with the original indicator values ​​to obtain the performance change.

[0194] The control parameters for a dual-motor switching pure electric bus cannot simply pursue the optimal values ​​in a single simulation; they also need to maintain stability under different driver operations, road disturbances, and device deviations. This invention not only dynamically adjusts the weights of the four objectives according to the operating conditions during the search phase, but also introduces a secondary screening mechanism on the Pareto optimal candidate set in the final decision-making phase, further considering robustness under parameter perturbations. The final optimal cooperative control parameter vector obtained through this method... It can simultaneously take into account the response speed of mode switching, the smoothness of switching, the overall efficiency of dual motors and the loss of motor life, so as to achieve rapid response and smooth coordinated drive control of pure electric buses during mode switching.

[0195] S5. Application of Dual-Motor Fast Drive Cooperative Control Based on Operating Condition Adaptation

[0196] Based on the decision results of the working condition sensitive area constructed by the aforementioned steps, the improved gray wolf optimization algorithm, the multi-objective comprehensive fitness function, and the Pareto optimal candidate set, this invention ultimately forms a complete application scheme for rapid drive cooperative control of pure electric buses.

[0197] In the offline phase, the processes described in S2 to S4 are first executed. Using the collected and labeled historical operating data, the improved Grey Wolf optimization algorithm is used to iteratively optimize and obtain multiple optimal cooperative control parameter vectors covering different operating conditions. Then, a parameter-operating condition matching library is constructed using the operating condition feature vectors as indices. In this matching library, each typical operating condition (such as start-up, acceleration, cruising, and hill climbing) corresponds to one or more candidate control parameter vectors that perform well in robustness evaluation.

[0198] During actual vehicle operation, the vehicle controller receives real-time vehicle status signals from the CAN bus, including current vehicle speed, longitudinal acceleration, road gradient, battery state of charge, and real-time torque and speed of the dual motors. These signals are combined to form a current operating condition feature vector. Then, the controller calls the operating condition identification module to calculate the similarity between the current operating condition feature vector and the feature vectors of typical operating condition templates in the parameter-operating condition matching library, thereby identifying the current operating condition type of the vehicle.

[0199] Based on the identification results of the operating condition type, the controller retrieves the optimal coordinated control parameter vector with the highest matching degree from the matching library, which serves as the control benchmark for the dual-motor drive system at the current moment. During rapid driving, especially when mode switching is triggered, the controller calculates key control quantities such as torque distribution coefficient, mode switching threshold, proportional-integral-derivative control gain, synchronization time window, and torque compensation coefficient in real time based on the retrieved optimal parameter vector. These control quantities are sent to the first motor controller and the second motor controller respectively, dynamically adjusting the output torque and operating mode of the dual motors. This ensures that upon receiving the driver's acceleration intention or the vehicle energy management strategy switching command, the dual-motor system can smoothly transition between single-motor mode and dual-motor mode with the shortest response time, minimal impact, highest overall efficiency, and minimal lifespan loss, or achieve rapid and precise coordinated torque distribution in dual-motor mode. In addition, the controller also has online fine-tuning capabilities. When it is detected that the operating conditions change continuously in a short period of time or that a transitional operating condition not fully covered in the matching library occurs, the system can perform fast interpolation or local online optimization based on a small neighborhood near the current optimal solution to generate control parameters that are adapted to the current transient process, thereby further enhancing the operating condition adaptability of the control system.

[0200] By combining offline optimization with online application, this invention achieves a dual-motor drive collaborative control effect for pure electric buses under different driving conditions, characterized by rapid response, smooth switching, high efficiency and energy saving, while also ensuring reliability.

[0201] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A method for coordinated control of dual motors in an electric bus, characterized in that, The method includes: Acquire multi-source state data of pure electric buses during operation and construct operating condition feature vectors; Based on the operating condition feature vector, perform operating condition clustering analysis on historical operating data to construct a mapping relationship between operating condition features and collaborative control parameters; A dual-motor cooperative control parameter optimization model is constructed, and the cooperative control parameters are optimized under the constraints of motor efficiency spectrum, battery power boundary and system stability to obtain the optimal cooperative control parameter vector; Establish a parameter-operating condition matching library, and identify the operating condition based on the current operating condition feature vector, and retrieve the corresponding optimal cooperative control parameter vector from the matching library; Based on the optimal cooperative control parameter vector, the torque distribution coefficient, mode switching threshold and control gain parameter are calculated in real time, and the first motor and the second motor are controlled to perform cooperative driving and mode switching.

2. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The constructed working condition feature vector includes: Vehicle operation data is collected through the vehicle's CAN bus, on-board sensors, and motor controller of the pure electric bus. The operating data includes at least vehicle speed, longitudinal acceleration, road gradient, battery state of charge, real-time torque of the first motor, real-time torque of the second motor, real-time speed of the first motor, and real-time speed of the second motor. The operating data is collected synchronously at a fixed sampling frequency and time-aligned to form vehicle operating status data corresponding to the same moment. Based on the vehicle operating status data, a condition feature vector is constructed to characterize the current driving state of the pure electric bus.

3. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The steps for constructing the mapping relationship between operating condition characteristics and cooperative control parameters include: The collected historical operating data of pure electric buses were processed by time alignment, outlier removal and normalization. Using the operating condition feature vector as input, the K-means clustering algorithm is used to perform cluster analysis on historical operating data, and the number of operating condition clusters is determined by the elbow rule. Multiple operating condition clusters corresponding to typical operating states of pure electric buses are obtained. Each operating condition cluster corresponds to starting operating condition, acceleration operating condition, cruising operating condition, climbing operating condition or frequently switching operating condition. For each operating condition cluster, its central eigenvector and covariance matrix are calculated to characterize the distribution characteristics of the vehicle's operating state under that operating condition. Based on the distance relationship between the current working condition feature vector and each working condition cluster, the degree of membership of the current working condition to each working condition cluster is calculated.

4. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The construction of the dual-motor cooperative control parameter optimization model includes: Define a dual-motor cooperative control parameter vector, which includes torque distribution coefficient, mode switching threshold, proportional control gain, integral control gain, derivative control gain, speed switching threshold, synchronization time window, and torque compensation coefficient. Based on the current operating condition characteristic vector of the pure electric bus, combined with the efficiency spectrum, output torque range and battery power boundary of the first and second motors, the value range of each control parameter is determined respectively. The value range is a dynamic upper and lower bound that changes with the vehicle's operating state, used to limit the feasible value range of the cooperative control parameters under the physical constraints of the dual-motor drive system.

5. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The multi-objective optimization includes: The optimization targets are mode switching response time, mode switching impact, overall efficiency of dual-motor system, and motor life loss. Under the premise of satisfying the motor output torque limit, battery power limit and system stability constraints, the cooperative control parameter vector is optimized and solved; The optimization objectives are used to characterize the power performance, smoothness, energy efficiency, and reliability of pure electric buses during dual-motor drive and mode switching processes.

6. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The optimization model adopts the Grey Wolf optimization algorithm. During the population initialization phase, an initial population is generated based on the distribution characteristics of the cooperative control parameters under different operating conditions. During the iteration process, the search for cooperative control parameters is achieved by updating the individual positions of gray wolves; During the search process, the individual components that do not meet the constraints are corrected by combining the motor efficiency graph and torque constraints. The optimal cooperative control parameter vector that meets the requirements of multi-objective optimization is obtained through iterative updates.

7. The method for coordinated control of dual motors in an electric bus according to claim 1, characterized in that, The construction and use of the parameter-condition matching library includes, The optimal collaborative control parameter vectors corresponding to each operating condition cluster are stored to form a parameter-operating condition mapping relationship; During the operation of a pure electric bus, its membership degree to each operating condition cluster is calculated based on the current operating condition feature vector. Select the cooperative control parameter vector corresponding to the operating condition cluster with the highest degree of membership as the current control parameter; The control parameters are input into the vehicle controller to control the torque output and operating mode switching of the first and second motors.

8. A dual-motor cooperative control system for an electric bus, characterized in that, The system includes: The data acquisition module is used to acquire vehicle operating status data; The operating condition identification module is used to construct operating condition feature vectors and identify the current operating condition; The parameter optimization module is used to build a collaborative control parameter optimization model based on historical data and generate optimal parameters. The parameter matching module is used to build a parameter-condition matching library and retrieve the optimal coordinated control parameters; The control execution module is used to control the output torque and operating mode of the two motors according to the optimal cooperative control parameters.

9. An electronic device, characterized in that, Including processor and memory, The memory stores a computer program, and when the processor executes the computer program, it implements the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-7.

Images