Pure electric commercial vehicle drive execution layer control and multi-motor torque optimization distribution method

Abstract

The present application belongs to the technical field of energy-saving control of electric drive system, and in particular to a driving execution layer control and multi-motor torque optimization distribution method for pure electric commercial vehicles. 2 The method comprises the following steps: S1: calculating total loss; S2: using a method combining a trend awareness mechanism and a golden section search to perform minimum loss control; S3: establishing a multi-motor torque distribution model; S4: using E An optimization algorithm is used to solve the multi-motor torque distribution coefficient; and S5: issuing a target torque and forming a closed-loop control. The present application establishes a driving permanent magnet synchronous motor loss model, takes copper loss and iron loss as main optimization objects, and provides a clear objective function basis for minimum loss control, which is conducive to directly reducing motor operating loss from the driving execution layer; and a trend awareness mechanism TAM is introduced on the basis of the golden section search method, so as to improve the search efficiency, convergence speed and working condition adaptability of the minimum loss control.

Description

Technical Field

[0001] This invention relates to the field of energy-saving control technology for electric drive systems, specifically to a method for drive execution layer control and multi-motor torque optimization distribution for pure electric commercial vehicles. Background Technology

[0002] Pure electric commercial vehicles offer advantages such as zero emissions, low noise, and high transmission efficiency, and are widely used in urban logistics, sanitation transportation, and short-distance shuttle services. Compared to passenger vehicles, commercial vehicles are heavier, operate under more complex conditions, and experience more frequent start-stop cycles, resulting in higher overall energy consumption and a larger proportion of energy costs in their operating expenses. Due to limitations in battery energy density, the driving range of pure electric commercial vehicles remains limited. Therefore, reducing overall vehicle energy consumption and increasing driving range while maintaining basic vehicle power and operational performance has significant engineering application value.

[0003] In existing pure electric commercial vehicle drive control, the drive motor controller and torque distribution module typically execute the control separately after the upper layer issues a target drive request. For efficient drive motor operation, existing minimum loss control methods often employ fixed-step search or conventional iterative strategies, which suffer from sensitivity to initial conditions, low search efficiency, and insufficient online adaptability, making it difficult to simultaneously achieve control accuracy and real-time performance under complex operating conditions. Furthermore, while existing golden section search methods are suitable for solving the optimal value in a single-peak interval, they still suffer from drawbacks such as being prone to getting trapped in local optima, lacking memory of historical information, and being sensitive to the initial interval.

[0004] On the other hand, in multi-motor drive scenarios, torque distribution typically needs to minimize the total power consumption of the four drive permanent magnet synchronous motors while meeting the overall vehicle drive requirements and the constraints of each wheel. Existing torque distribution methods often suffer from high computational complexity, slow convergence speed, or insufficient adaptability to complex operating conditions when facing scenarios with nonlinear efficiency characteristics, numerous constraints, and high real-time requirements, making it difficult to quickly obtain the optimal distribution result that balances energy efficiency and execution feasibility. To address these issues, this invention employs E... 2 An optimized algorithm is used to solve the torque distribution coefficient of multiple motors, thereby improving the accuracy and real-time performance of the solution.

[0005] Therefore, it is necessary to propose an energy efficiency optimization control method for the drive execution layer of pure electric commercial vehicles. After receiving the target drive request from the upper layer, the method can improve the energy efficiency optimization capability, online solution speed and adaptability of the drive system under complex working conditions through the coordinated processing of minimum loss control of the drive permanent magnet synchronous motor and torque optimization distribution of multiple motors. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0007] To address the aforementioned technical problems, according to one aspect of the present invention, the present invention provides the following technical solution: a method for drive execution layer control and multi-motor torque optimization allocation for pure electric commercial vehicles, comprising the following steps:

[0008] S1: Establish models for copper loss, iron loss, mechanical loss and other losses of the permanent magnet synchronous motor, and calculate the total loss accordingly;

[0009] S2: A method combining trend-aware mechanisms and golden ratio search is used for minimum loss control: historical search trajectories are analyzed to determine the effectiveness of the current golden ratio search trend, and the search range is intelligently adjusted accordingly. The position and size of the golden ratio guide the more likely areas for a precise search.

[0010] S3: Establish a multi-motor torque distribution model: Torque distribution mainly considers obtaining a set of torque distribution coefficients to minimize the total power consumption of the four driving permanent magnet synchronous motors;

[0011] S4: Using E 2 Optimization algorithm for solving multi-motor torque distribution coefficients: The torque distribution coefficients are... As the vector to be optimized, using E 2 The optimization algorithm performs torque allocation optimization, and the fitness function... as follows:

[0012] Randomly generated in the search space There are 1 feasible solutions, and the position of each solution is determined by a... The dimensional vector representation means that the solution iteratively updates its position until a preset maximum number of iterations is satisfied. Through this E 2 Algorithm optimizes torque distribution coefficient ;

[0013] S5: Issue the target torque and form a closed-loop control.

[0014] As a preferred embodiment of the pure electric commercial vehicle drive execution layer control and multi-motor torque optimization allocation method described in this invention, wherein the copper loss in S1 The expression is: ,in and These are the d-axis and q-axis currents of the permanent magnet synchronous motor, respectively. Stator resistance;

[0015] The iron loss Including hysteresis loss Eddy current loss and additional losses The iron loss By introducing equivalent iron loss resistance To approximate the calculation, the stator current has the following relationship after introducing the iron loss resistance:

[0016]

[0017] in, These are the iron loss currents along the d-axis and q-axis. For the torque currents along the d-axis and q-axis;

[0018] When the motor is running stably, the iron loss current Represented as:

[0019]

[0020] The electric angular velocity of the motor. , For direct-axis and quadrature-axis inductors, For permanent magnet flux linkage;

[0021] Formula for calculating iron loss power:

[0022] ;

[0023] The mechanical wear Including bearing friction loss Rotor aerodynamic losses The calculation formula is as follows:

[0024]

[0025] in This is the bearing load factor. For the bearing diameter, The surface roughness coefficient of the rotor is . The coefficient of friction, air density, The rotor radius is... The axial length of the rotor This refers to the mechanical angular velocity of the motor.

[0026] The total loss Represented as:

[0027] .

[0028] As a preferred embodiment of the pure electric commercial vehicle drive execution layer control and multi-motor torque optimization allocation method described in this invention, the combination point of the trend perception mechanism and the golden section search in S2 is as follows:

[0029] First, define the GS search range. midpoint As the representative state for the current iteration number, the midpoint value is... The objective function value corresponding to the midpoint Stored separately in the TAM mechanism and In the middle, initialize the GS and TAM parameters and set the iteration threshold. ;

[0030] Calculate the trend vector using the following formula :

[0031]

[0032] when When, it indicates that the nearest point is better; when When this occurs, it indicates that a better solution exists in the trend direction. Through TAM intervention, GS no longer relies entirely on the golden ratio scaling, but instead adaptively compresses the interval boundaries along the favorable trend direction. The compression step size is determined by a coefficient. The control is calculated using the following formula:

[0033] .

[0034] As a preferred embodiment of the pure electric commercial vehicle drive execution layer control and multi-motor torque optimization allocation method described in this invention, in step S3, for the drive motor, its power consumption... Represented as:

[0035]

[0036] in, The efficiency of the four motors is expressed as... and The function, with Obtained by looking up a table in the form of a graph;

[0037] The objective function for torque distribution is:

[0038]

[0039] The constraints are:

[0040]

[0041] The torque distribution problem is to find a set of optimal four-wheel torques. , so that the objective function To achieve the minimum, define the torque distribution coefficient. Represent Its expression is as follows:

[0042] .

[0043] As a preferred embodiment of the pure electric commercial vehicle drive execution layer control and multi-motor torque optimization allocation method described in this invention, wherein in step S4, during iteration... Next time, the solution The position is represented as In addition to location, each solution also has a memory to store the best locations it has experienced during iteration. Next time, the solution The memory location is represented as ;

[0044] To update its position, solve It will randomly choose to use either the "explore" or "exploit" strategy. If "explore" is chosen, its position will be updated to a random position within the search space; if "exploit" is chosen, it will first be based on a solution randomly selected from the population. memory location To solve Generate a hypothetical location The formula is as follows:

[0045]

[0046] in Representing a random number between 0 and 1, the distance between the generated hypothetical position and the memory position of the selected solution is calculated; this distance determines the iteration... Next time solution Utilization radius The formula is as follows:

[0047]

[0048] After determining the radius, the solution is obtained using the pattern defined by the following formula. In dimensions Final position on:

[0049]

[0050] in, It is a random input parameter, for each dimension of the new position. The values ​​are all generated independently.

[0051] As a preferred embodiment of the pure electric commercial vehicle drive execution layer control and multi-motor torque optimization allocation method described in this invention, the specific method of S5 is as follows: the target torque of each drive motor obtained in S4 is sent down to the motor drive execution unit for execution to form the longitudinal driving force of the whole vehicle; at the same time, the vehicle speed, wheel speed, motor operating status, battery status and current energy efficiency status generated during vehicle operation are fed back in real time to the human-machine cooperative driving decision unit and the drive energy efficiency optimization unit for driver intention recognition, economic driving reference quantity update, cooperative weight allocation, minimum loss control and multi-motor torque allocation in the next control cycle, thereby forming a complete closed-loop control.

[0052] Compared with the prior art, the beneficial effects of the present invention are: the present invention establishes a loss model for driving permanent magnet synchronous motors, takes copper loss and iron loss as the main optimization objects, provides a clear objective function basis for minimum loss control, and is conducive to directly reducing motor operating losses from the drive execution layer.

[0053] This invention introduces a trend-aware mechanism (TAM) based on the golden section search method. By analyzing historical search trajectories, it determines whether the current search trend is effective and adaptively adjusts the position and size of the search interval accordingly. This makes the search process combine the global convergence characteristics of the golden section method with the historical memory and direction prediction capabilities of the trend-aware mechanism, thereby improving the search efficiency, convergence speed, and adaptability of minimum loss control.

[0054] This invention establishes a torque distribution model with the objective of minimizing the total power consumption of the four driving permanent magnet synchronous motors, and uses E 2 The optimization algorithm solves for the torque distribution coefficient, which can improve the response accuracy and solution speed of multi-motor torque distribution and enhance the adaptability to complex nonlinear working conditions.

[0055] This invention incorporates minimum loss control and multi-motor torque distribution into the same drive execution layer energy efficiency optimization chain. After receiving the target drive request from the upper layer, the efficient operation control quantity of the drive motor can be solved first, and then the target torque of each wheel can be obtained, thereby improving the energy efficiency, smoothness and feasibility of the conversion process from the target drive request to the wheel-end execution result.

[0056] The average power loss is reduced by about 33W after the introduction of the golden section search method in this invention. The power loss is further reduced by about 10W after the introduction of the search method combining TAM and golden section. The response time is also shortened from about 0.04s to about 0.02s. This shows that the method proposed in this invention has improved both the power loss optimization effect and the dynamic response performance. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0058] Figure 1 This is the equivalent circuit diagram of the d-axis and q-axis of the permanent magnet synchronous motor of the present invention;

[0059] Figure 2 This is a flowchart of the TAM process of the present invention. Detailed Implementation

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0061] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0062] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0063] This invention provides a method for drive execution layer control and multi-motor torque optimization allocation for pure electric commercial vehicles, comprising the following steps:

[0064] Step 1: Establish a loss model for driving the permanent magnet synchronous motor:

[0065] As a core component for converting electrical energy into mechanical energy, the PMSM inevitably generates various losses during operation: copper losses caused by the resistance of the stator windings, hysteresis and eddy current iron losses induced by the alternating magnetic field in the stator core, mechanical losses caused by bearing friction and wind resistance, and a small amount of stray losses that are difficult to model. Among these, copper losses and iron losses are controllable losses, while mechanical losses and other losses are uncontrollable losses.

[0066] Copper loss: Copper loss of PMSM Copper losses are generated by the stator current flowing through the motor windings and are mainly divided into two categories: fundamental frequency copper losses and high-frequency additional copper losses. Fundamental frequency copper losses originate from the Joule heat generated by the phase current flowing through the stator winding resistance; while high-frequency additional copper losses are extra energy losses caused by the skin effect and proximity effect under alternating current. Considering the low operating frequency of hub-driven PMSMs, high-frequency copper losses can be ignored, and the result is simplified to the following expression: It can be seen that optimization and It can effectively reduce the copper loss of PMSM. and These are the d-axis and q-axis currents of the permanent magnet synchronous motor, respectively. This is the stator resistance.

[0067] Iron loss: Iron loss of PMSM Mainly composed of hysteresis loss Eddy current loss and additional losses Composition. Hysteresis loss originates from the displacement and flipping of magnetic domains within the permanent magnet material under the action of an alternating magnetic field, resulting in energy dissipation caused by friction between the domains. Its magnitude is mainly affected by the amplitude of the magnetic flux density and the frequency of magnetic field changes. Eddy current loss is caused by eddy currents induced in the permanent magnet by a periodically changing magnetic field, leading to Joule heating. This loss is closely related to the magnetic field strength, frequency of change, and waveform. Abnormal loss is mainly caused by factors such as uneven magnetic flux density distribution, local magnetic flux fluctuations, and irregular eddy current paths, manifesting as additional energy loss exceeding the predictions of classical hysteresis and eddy current models. Based on the Bertotti iron loss model, the following formula for calculating iron loss can be derived: in This is the hysteresis loss coefficient. This is the eddy current loss coefficient. For additional loss coefficient, The magnetic flux density frequency, The amplitude of the sinusoidal magnetic flux density. These are empirical coefficients. It can be seen that the Bertotteti iron loss model involves numerous parameters and complex calculations, making it unsuitable for real-time control of PMSM. Therefore, this invention introduces an equivalent iron loss resistance. The method is used to approximate the iron loss of the PMSM. The d-axis and q-axis equivalent circuits of the PMSM are as follows: Figure 1 As shown. Among them, These are the iron loss currents along the d-axis and q-axis. For the torque current along the d-axis and q-axis, Let be the equivalent resistance of the iron loss. After introducing the iron loss resistance, the stator current exhibits the following relationship:

[0068]

[0069] When the motor is running stably, the iron loss current It can be represented as:

[0070]

[0071] The electric angular velocity of the motor. , For inductance along the direct axis (d-axis) and quadrature axis (q-axis), It is a permanent magnet flux linkage.

[0072] PMSM iron loss power calculation formula:

[0073]

[0074] It can also be seen that through reasonable regulation and It can directly affect the iron loss power of PMSM.

[0075] Mechanical loss: Mechanical loss of PMSM Mainly caused by bearing friction loss Rotor aerodynamic losses Composition. Bearing friction loss is mainly generated by the contact friction between the rotor shaft and the internal balls and inner and outer rings of the bearing during motor operation. This loss is affected by various factors, including the load on the bearing, the diameter of the balls, the quality of lubrication, and the rotational speed of the motor. Rotor aerodynamic loss is generated by the interaction between the rotor and the surrounding air during high-speed rotation. Its magnitude depends on the roughness of the rotor surface, the flow field characteristics of the air gap region, the structural characteristics of the rotor itself, as well as the operating speed and the viscosity characteristics of the air. The calculation formula is as follows:

[0076]

[0077] in This is the bearing load factor. For the bearing diameter, The surface roughness coefficient of the rotor is . The coefficient of friction, air density, The rotor radius is... The axial length of the rotor. This refers to the mechanical angular velocity of the motor.

[0078] It can be seen that the mechanical losses of the PMSM are only related to the rotor electric angular velocity and the motor structural parameters. Adjusting the load or and The magnitude of mechanical wear cannot be changed.

[0079] Other losses: Other losses of PMSM Mainly due to stator stray losses With rotor stray losses The system consists of two parts. Stator stray losses are mainly caused by the leakage magnetic field and harmonic magnetic field generated by the stator current in the air gap. Their magnitude is affected by factors such as the stator winding structure, current harmonic content, slot opening effect, and the magnetic permeability of the material. Rotor stray losses originate from high-order spatial harmonics and high-frequency eddy currents excited by the cogging effect in the air gap magnetic field. Their value is closely related to the rotor structure, air gap magnetic flux density distribution, power supply quality, and motor operating conditions. Due to the high complexity of modeling other losses, it is difficult to optimize them through real-time control.

[0080] In summary, among the various losses during PMSM operation, the main optimization objectives of the LMC strategy are copper loss and iron loss. Substitution and combined and The total loss of PMSM can be obtained. It can be represented as:

[0081] .

[0082] Step 2: Minimum loss control is achieved by combining a trend-aware mechanism with the golden ratio search.

[0083] Golden Section Search Algorithm:

[0084] When the motor is running in steady state Not explicit With rigorous proof It is about The concave function, and the logic of minimum loss control, is to solve for a set of optimal d-axis currents. , making To reach the minimum. The golden section method is suitable for solving intervals. The basic principle for finding the minimum value of any concave function on the upper bound is as follows:

[0085] Let the concave function In the interval There is a minimum value above. Take two points upon taking office and And satisfy If during the search ,but The minimum value is at Up; Conversely, if ,but The minimum value is at The algorithm divides the interval into three segments by inserting two points and comparing the function values ​​at these points. This allows for the removal of the leftmost or rightmost segment of the interval, based on the characteristics of a concave function, thus narrowing the search interval. This process is repeated, inserting points into new intervals to continuously shrink the search interval until the required accuracy is met. It's important to note that the insertion method of two points within the interval determines the convergence speed of the search algorithm. The purpose of the golden section method is to maximize the interval shrinkage speed by rationally determining the insertion point positions. Its golden section ratio... .

[0086] TAM mechanism:

[0087] 1. Overview of TAM Mechanism

[0088] Trend Awareness (TAM) is an innovative metaheuristic algorithm enhancement strategy. Its core idea is to dynamically guide the search direction of the optimization process by analyzing historical search trajectories. This mechanism solves the problem of insufficient utilization of historical position data in traditional optimization algorithms. By mining the movement trends of the population in continuous iterations, it significantly improves the search efficiency and accuracy of the algorithm.

[0089] The TAM mechanism relies on a historical search position matrix, which stores the position information of each individual in the population during each iteration, and a historical fitness matrix, which records the fitness value of the corresponding individual in each iteration.

[0090] The core components of TAM include trendline calculation, adaptive covariance mechanism, and fitness-based update strategy. The adaptive covariance mechanism balances historical trends with stochastic exploration by generating high-dimensional random vectors; its mathematical expression is as follows:

[0091]

[0092] in, Represents the historical movement direction vector. It is a random vector sampled from the covariance matrix estimated from the current population distribution. and These are dynamically adjusted weight parameters used to control the balance between historical trends and random exploration. This mechanism ensures that the algorithm tends to explore globally in the early stages of iteration, while focusing on fine-grained local search in later stages.

[0093] 2. Detailed explanation of TAM workflow (e.g.) Figure 2 (As shown)

[0094] Distance calculation:

[0095] TAM is based on the position points of each individual's two most recent iterations. and points Construct a trend line and calculate the trend line for any point in the historical data. Calculate the Euclidean distance to the trend line and select the closest one. Each point is used as an evaluation candidate set. Distance The calculation formula is:

[0096]

[0097] Adaptive covariance mechanism:

[0098] TAM introduces an adaptive covariance-based method to dynamically adjust the offset vector according to the current population distribution. This improves search efficiency and accuracy.

[0099] First, calculate the trend vector pointing in the direction of the historical position. :

[0100]

[0101] Then, an adaptive covariance mechanism is used to generate high-dimensional vectors. ,

[0102]

[0103] in, Represents the historical movement direction vector. Represents the offset vector. It is a Gaussian random vector with zero mean. and These are dynamically adjusted weighting parameters used to control the balance between historical trends and random exploration. Calculated based on the current population location, where The average position of all individuals across all dimensions:

[0104]

[0105] To ensure amplitude and Equal, for Normalization is performed:

[0106] .

[0107] Location update:

[0108] TAM for the selected fitness of neighboring points Comparison, among which , According to the following formula Assign a value:

[0109]

[0110] like This indicates that neighboring points are better, and individuals will move away from the trend direction; if This indicates that there is a better solution in the trend direction, and the individual will move in the trend direction:

[0111] .

[0112] Combining the TAM mechanism with the golden section search method:

[0113] The golden section search method has drawbacks such as being prone to getting trapped in local optima, having no memory of historical information, and being sensitive to the initial interval. To address these shortcomings, this invention employs the TAM mechanism, which analyzes historical search trajectories to determine the effectiveness of the current golden section search trend and intelligently adjusts the search interval accordingly. The location and size of the area determine the direction of the golden ratio, thus guiding a more precise search within the most likely region. Its main points of convergence are as follows:

[0114] First, define the GS search range. midpoint As the representative state for the current iteration number, the midpoint value is... The objective function value corresponding to the midpoint Stored separately in the TAM mechanism and In the middle, initialize the GS and TAM parameters and set the iteration threshold. .

[0115] Calculate the trend vector using the following formula :

[0116]

[0117] And it is updated via the TAM mechanism described above, when When, it indicates that the nearest point is better; when This indicates that a better solution exists in the trend direction. With the intervention of TAM, GS no longer relies entirely on the golden ratio scaling, but instead adaptively compresses the interval boundaries along the favorable trend direction. The compression step size is determined by a coefficient. The control is calculated using the following formula:

[0118]

[0119] In summary, the TAGS proposed in this invention successfully achieves the complementary advantages of two types of methods: on the one hand, it inherits the global convergence guarantee of the golden section method in single-peak problems, and on the other hand, it introduces the historical perception capability of the TAM mechanism, enabling the search process to have the characteristics of "intelligent memory" and "direction prediction".

[0120] Step 3: Establish a multi-motor torque distribution model:

[0121] Torque distribution primarily considers obtaining a set of torque distribution coefficients that minimizes the total power consumption of the four drive permanent magnet synchronous motors. For each drive motor, the power consumption is... It can be represented as:

[0122]

[0123] in, This indicates the efficiency of the four motors. The output torque of the four motors is about and The function, with Obtained by looking up a table in the form of a graph.

[0124] The objective function for torque distribution is:

[0125]

[0126] The constraints are:

[0127]

[0128] The torque distribution problem is to find a set of optimal four-wheel torques. , so that the objective function To achieve the minimum, define the torque distribution coefficient. Represent Its expression is as follows:

[0129] .

[0130] Step 4: Use E 2 Optimization algorithm to solve for the torque distribution coefficient of multiple motors:

[0131] Torque distribution coefficient As the vector to be optimized, using E 2 The optimization algorithm performs torque distribution optimization, and its fitness function... As shown below:

[0132] E 2 It is a population-based optimization method. First, random generation is performed in the search space. There are 1 feasible solutions. The position of each solution is determined by a... The solution is represented by a 3D vector. The solution iteratively updates its position until a preset maximum number of iterations is satisfied. In iteration Next time, the solution The position is represented as In addition to location, each solution also possesses a memory that stores the best locations it has experienced. During iteration... Next time, the solution The memory location is represented as .

[0133] To update its position, solve It will randomly choose to use either the "explore" or "exploit" strategy. If "explore" is chosen, its position will be updated to a random position within the search space; if "exploit" is chosen, it will first be based on a solution randomly selected from the population. memory location To solve Generate a hypothetical location The formula is as follows:

[0134]

[0135] in This represents a random number between 0 and 1. Then, the distance between the generated hypothetical position and the memory position of the selected solution is calculated. This distance determines the position in the iteration. Next time solution Utilization radius The formula is as follows:

[0136]

[0137] After determining the radius, the solution is obtained using the pattern defined by the following formula. In dimensions Final position on:

[0138]

[0139] in, It is a random input parameter, for each dimension of the new position. The values ​​are all uniquely generated. The update method described above is used iteratively until the maximum number of iterations is reached, and then the value of E is determined. 2 Algorithm optimizes torque distribution coefficient .

[0140] Step 5: Issue the target torque and form a closed-loop control

[0141] The target torques of each drive motor obtained in step 4 are sent to the motor drive execution unit for execution, forming the longitudinal driving force of the whole vehicle. At the same time, the vehicle speed, wheel speed, motor operating status, battery status and current energy efficiency status generated during vehicle operation are fed back to the human-machine cooperative driving decision unit and drive energy efficiency optimization unit in real time. These are used for driver intention recognition, economic drive reference quantity update, cooperative weight allocation, minimum loss control and multi-motor torque allocation in the next control cycle, thus forming a complete closed-loop control.

[0142] Example 1: Implementation of a Driver Execution Layer Based on TAM-Golden Section Minimum Loss Control

[0143] An energy efficiency optimization control method for the drive execution layer of a pure electric commercial vehicle is disclosed, applied to a four-wheel independent wheel-end drive pure electric commercial vehicle. The vehicle is equipped with four wheel-end drive permanent magnet synchronous motors. The drive execution layer receives the target drive request output from the upper-level control module and solves for the efficient operation control quantities of the drive motors based on a motor loss model and a minimum loss control algorithm. During operation, the drive permanent magnet synchronous motors generate copper losses, iron losses, mechanical losses, and other losses. Copper losses and iron losses are the main optimization targets for minimum loss control, while mechanical losses and other losses are included as components of the total losses in loss estimation.

[0144] In this embodiment, a loss model for driving the permanent magnet synchronous motor is first established. Copper losses are generated by the stator current through the winding resistance, iron losses are approximated by introducing an equivalent iron loss resistance, mechanical losses consist of bearing friction losses and rotor aerodynamic losses, and other losses are mainly stator stray losses and rotor stray losses. Based on the above model, a total loss expression for driving the permanent magnet synchronous motor is constructed for subsequent minimum loss control optimization.

[0145] Furthermore, under steady-state operation conditions of the drive motor, the golden section search method is used to search for the control variable that minimizes the total loss within a given interval. To overcome the shortcomings of traditional golden section search, such as lack of memory for historical information and susceptibility to the influence of the initial interval, a trend-aware mechanism (TAM) is introduced in this embodiment. The TAM mechanism analyzes the search trajectory in continuous iterations through the historical search position matrix and historical fitness matrix, constructs a trend line, selects several nearest neighbor points as the evaluation candidate set, and uses an adaptive covariance mechanism to generate an offset vector to balance historical trends and random exploration.

[0146] In this embodiment, the midpoint of the golden section search interval is taken as the current iteration state, and the midpoint value and the corresponding objective function value are written into the historical matrix of the TAM mechanism. Then, based on the trend vector and the fitness comparison results of neighboring points, it is determined whether the current search trend is effective. When neighboring points are better, the search interval is updated away from the current trend direction; when a better solution exists in the trend direction, the search interval is adaptively compressed along the favorable direction. In this way, the golden section method no longer relies entirely on fixed-ratio scaling, but performs a refined search under the guidance of a trend-aware mechanism, thereby achieving faster convergence speed and higher optimal search accuracy.

[0147] In this embodiment, simulations were conducted using a conventional control method, a fixed convergence rate golden section search method, and a TAM-optimized golden section search method for comparison. The results show that the average power loss is reduced by approximately 33W after introducing the golden section search method, and further reduced by approximately 10W after introducing the search method combining TAM and the golden section. Furthermore, the response time is shortened from approximately 0.04s to approximately 0.02s, indicating that the minimum loss control method proposed in this invention is superior to the comparative methods in both loss reduction and dynamic response.

[0148] Example 2: Based on E 2 Implementation Examples of Multi-Motor Torque Distribution Optimization Algorithms

[0149] In another embodiment, after receiving the total drive demand from the upper layer, the drive execution layer further performs four-wheel drive torque distribution. In this embodiment, the torque distribution model is established with the goal of minimizing the total power consumption of the four drive permanent magnet synchronous motors. For each drive motor, its power consumption is determined by the motor efficiency, which is a function of speed and torque and can be obtained by looking up a table using an efficiency chart. Based on this, a target function for the total power consumption of the four motors is constructed, and corresponding constraints are set to solve for the optimal four-wheel target torque.

[0150] In this embodiment, a torque distribution coefficient is defined as a variable to be optimized, used to characterize the distribution ratio of the four drive motors in the total drive request. Then, the torque distribution coefficients are used to construct an optimization vector, employing E... 2 The solution is obtained using an optimization algorithm. The E... 2 The optimization algorithm is a population-based optimization method. First, several feasible solutions are randomly generated in the search space. Each solution is represented by a position vector, and its historical best memory position is retained. During the iteration process, each solution randomly updates its position using either an "exploration" or "exploitation" strategy. The "exploration" strategy is used to enhance the global search capability, while the "exploitation" strategy generates a hypothetical position based on the memory positions of other individuals and performs local optimization search based on the utilization radius.

[0151] As the iteration progresses, E 2The algorithm continuously corrects the torque distribution coefficients by updating the memory position, utilization radius, and random perturbations in each dimension until a preset maximum number of iterations is reached, ultimately outputting the optimal torque distribution result for the four drive motors. This implementation method improves the torque distribution solution speed and response accuracy while reducing the total power consumption of the four drive permanent magnet synchronous motors, all while meeting the overall vehicle drive requirements and wheel constraints.

[0152] Example 3: Integrated Energy Efficiency Optimization Example for Driver Execution Layer

[0153] In another embodiment, the drive execution layer implements minimum loss control in series with multi-motor torque distribution. Specifically: first, it receives the target drive request output from the upper-level control module; second, based on the loss model of the permanent magnet synchronous motor and the TAM-golden section minimum loss control method, it solves for the efficient operation control quantities of the drive motor; then, based on this, it establishes a torque distribution model with the objective of minimizing the total power consumption of the four motors, and uses E... 2 The optimization algorithm solves for the torque distribution coefficient of the four-wheel drive, and finally obtains the target torque commands for the left front wheel, right front wheel, left rear wheel and right rear wheel, and sends the target torque commands to the motor drive execution unit.

[0154] In this embodiment, the target torque of each wheel output by the drive execution layer not only satisfies the target drive request of the upper layer, but also takes into account the requirements of reducing motor losses and optimizing the overall vehicle drive energy efficiency, which is conducive to improving the energy efficiency, responsiveness and execution feasibility of pure electric commercial vehicles under complex working conditions.

[0155] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for drive execution layer control and multi-motor torque optimization allocation in pure electric commercial vehicles, characterized in that, Includes the following steps: S1: Establish models for copper loss, iron loss, mechanical loss and other losses of the permanent magnet synchronous motor, and calculate the total loss accordingly; S2: A method combining trend-aware mechanisms and golden ratio search is used for minimum loss control: historical search trajectories are analyzed to determine the effectiveness of the current golden ratio search trend, and the search range is intelligently adjusted accordingly. The position and size of the golden ratio guide the more likely areas for a precise search. The points of convergence between the trend perception mechanism and the golden ratio search are as follows: First, define the GS search range. midpoint As the representative state for the current iteration number, the midpoint value is... The objective function value corresponding to the midpoint Stored separately in the TAM mechanism and In the middle, initialize the GS and TAM parameters and set the iteration threshold. ; Calculate the trend vector using the following formula : when When, it indicates that the nearest point is better; when When this occurs, it indicates that a better solution exists in the trend direction. Through TAM intervention, GS no longer relies entirely on the golden ratio scaling, but instead adaptively compresses the interval boundaries along the favorable trend direction. The compression step size is determined by a coefficient. The control is calculated using the following formula: ； S3: Establish a multi-motor torque distribution model: Torque distribution mainly considers obtaining a set of torque distribution coefficients to minimize the total power consumption of the four driving permanent magnet synchronous motors; For the drive motor, its power consumption Represented as: in, The efficiency of the four motors is expressed as... and The function, with Obtained by looking up a table in the form of a graph; The objective function for torque distribution is: The constraints are: The torque distribution problem is to find a set of optimal four-wheel torques. , so that the objective function To achieve the minimum, define the torque distribution coefficient. Represent Its expression is as follows: ； S4: Using E 2 Optimization algorithm for solving multi-motor torque distribution coefficients: The torque distribution coefficients are... As the vector to be optimized, using E 2 The optimization algorithm performs torque allocation optimization, and the fitness function... as follows: Randomly generated in the search space There are 1 feasible solutions, and the position of each solution is determined by a... The dimensional vector representation means that the solution iteratively updates its position until a preset maximum number of iterations is satisfied. Through this E 2 Algorithm optimizes torque distribution coefficient ; S5: Issue the target torque and form a closed-loop control.

2. The method for drive execution layer control and multi-motor torque optimization allocation of pure electric commercial vehicles according to claim 1, characterized in that, copper loss in S1 The expression is: ,in and These are the d-axis and q-axis currents of the permanent magnet synchronous motor, respectively. Stator resistance; The iron loss Including hysteresis loss Eddy current loss and additional losses The iron loss By introducing equivalent iron loss resistance To approximate the calculation, the stator current has the following relationship after introducing the iron loss resistance: in, These are the iron loss currents along the d-axis and q-axis. For the torque currents along the d-axis and q-axis; When the motor is running stably, the iron loss current Represented as: The electric angular velocity of the motor. , For direct-axis and quadrature-axis inductors, For permanent magnet flux linkage; Formula for calculating iron loss power: ； mechanical wear Including bearing friction loss Rotor aerodynamic losses The calculation formula is as follows: in This is the bearing load factor. For the bearing diameter, The surface roughness coefficient of the rotor is . The coefficient of friction, air density, The rotor radius is... The axial length of the rotor This refers to the mechanical angular velocity of the motor. The total loss Represented as: 。 3. The method for drive execution layer control and multi-motor torque optimization allocation of pure electric commercial vehicles according to claim 1, characterized in that, In S4, during iteration Next time, the solution The position is represented as In addition to location, each solution also has a memory to store the best locations it has experienced during iteration. Next time, the solution The memory location is represented as ; To update its position, solve It will randomly choose to use either the "explore" or "exploit" strategy. If "explore" is chosen, its position will be updated to a random position within the search space; if "exploit" is chosen, it will first be based on a solution randomly selected from the population. memory location To solve Generate a hypothetical location The formula is as follows: in Representing a random number between 0 and 1, the distance between the generated hypothetical position and the memory position of the selected solution is calculated; this distance determines the iteration... Next time solution Utilization radius The formula is as follows: After determining the radius, the solution is obtained using the pattern defined by the following formula. In dimensions Final position on: in, It is a random input parameter, for each dimension of the new position. The values ​​are all generated independently.

4. The method for drive execution layer control and multi-motor torque optimization allocation of pure electric commercial vehicles according to claim 1, characterized in that, The specific method of S5 is as follows: the target torque of each drive motor obtained in S4 is sent to the motor drive execution unit for execution to form the longitudinal driving force of the whole vehicle; at the same time, the vehicle speed, wheel speed, motor operating status, battery status and current energy efficiency status generated during vehicle operation are fed back to the human-machine cooperative driving decision unit and the drive energy efficiency optimization unit in real time, which are used for driver intention recognition, economic driving reference quantity update, cooperative weight allocation, minimum loss control and multi-motor torque allocation in the next control cycle, thereby forming a complete closed-loop control.

Images