New energy vehicle permanent magnet synchronous motor optimization method based on adaptive evolutionary algorithm

Abstract

An optimization method for permanent magnet synchronous motors (PMSMs) in new energy vehicles based on an adaptive evolutionary algorithm includes: establishing a two-dimensional simulation model of the PMSM; constructing a multi-physics state-driven adaptive evolutionary algorithm, which, based on the Pangolin optimization algorithm, introduces a heuristic initialization population strategy based on physical sensitivity and employs adaptive weights driven by multi-physics states and frequency domain feature feedback calibration mutation; using the position and size parameters of the PMSM rotor magnets as optimization variables, iteratively optimizing using the adaptive evolutionary algorithm, and recording the objective function values ​​corresponding to different combinations of rotor magnet parameters; stopping the iteration when the iterative optimization reaches a preset condition, and outputting the position and size parameters of the rotor magnets corresponding to the recorded optimal objective function value as the optimization result. This invention can improve the optimization efficiency and accuracy of the motor while ensuring power output.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle technology, and specifically to an optimization method for permanent magnet synchronous motors in new energy vehicles based on an adaptive evolution algorithm. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs), with their advantages of high power density, high efficiency, and wide speed range, have become a core component of electric drive systems in new energy vehicles. However, in actual operating conditions, the built-in V-shaped magnet rotor structure easily leads to distortion of the air gap magnetic flux density waveform, generating higher-order spatial harmonics, which in turn cause significant cogging torque and torque pulsation. During vehicle operation, the performance requirements for the motor are extremely stringent, demanding not only good power output but also smooth and low-noise operation. Cogging torque and torque pulsation are the main excitation sources causing low-frequency electromagnetic noise in motors.

[0003] In existing technologies, besides utilizing the optimization algorithms built into traditional electromagnetic simulation software (such as particle swarm optimization, genetic algorithms, and aurora algorithms), new swarm intelligence optimization algorithms (such as improved elk herd optimization) have recently emerged for optimizing motor dimensions and structures. However, most of these existing algorithms focus on optimizing single stator-side structures (such as stator auxiliary slots), or face two major technical bottlenecks when dealing with multivariable, strongly coupled, and highly nonlinear electromagnetic field shaping problems such as rotor-embedded V-magnets: First, existing algorithms struggle to achieve a perfect dynamic balance between global wide-area exploration and local precise development in complex, multi-peaked solution spaces. When faced with extreme value transitions, they are still prone to getting trapped in local optima, resulting in limited convergence accuracy.

[0004] Secondly, existing multi-objective optimization frameworks often lack dynamic penalties and hard constraints on "average torque," which makes the algorithm prone to reducing torque ripple by drastically reducing magnetic flux during the optimization process. Ultimately, it cannot accurately find the best combination of motor rotor parameters that truly balances "ultimate noise reduction" and "strong power without attenuation," which severely limits the overall optimization efficiency and engineering practicality of the motor. Summary of the Invention

[0005] In view of this, the present invention provides an optimization method for permanent magnet synchronous motors of new energy vehicles based on an adaptive evolution algorithm, so as to improve the optimization efficiency and accuracy of the motor while ensuring power output.

[0006] An optimization method for permanent magnet synchronous motors in new energy vehicles based on an adaptive evolution algorithm includes: Step S1: Establish a two-dimensional simulation model of the permanent magnet synchronous motor for new energy vehicles; Step S2: Construct a multi-physics state-driven adaptive evolution algorithm. The multi-physics state-driven adaptive evolution algorithm is based on the pangolin optimization algorithm, introduces a heuristic initialization population strategy based on physical sensitivity, and adopts multi-physics state-driven adaptive weights and frequency domain feature feedback calibration mutation. Step S3: The position and size parameters of the permanent magnet synchronous motor rotor magnet in the two-dimensional simulation model established in step S1 are used as optimization variables. The multi-physics field state-driven adaptive evolution algorithm constructed in step S2 is used for iterative optimization, and the objective function values ​​corresponding to different combinations of rotor magnet parameters are recorded. Step S4: When the iterative optimization reaches the preset condition, stop the iteration and output the position parameters and size parameters of the rotor magnet corresponding to the recorded optimal objective function value as the optimization result.

[0007] The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm provided by the present invention has the following beneficial effects: (1) In the multiphysics-driven adaptive evolution algorithm constructed in this invention, a heuristic initialization population strategy based on physical sensitivity is introduced to replace the traditional blind chaotic or random mapping. This strategy establishes a local sensitivity model of the rotor V-shaped magnet parameters to electromagnetic torque pulsation and spatial harmonics. Driven by prior knowledge of the electromagnetic field, the initial population is concentrated in the parameter subspace where low-distortion waveforms are likely to be generated for adaptive sampling. This avoids wasting computing power in the degraded parameter region that is prone to generating high-frequency electromagnetic howling from the physical level, and greatly improves the engineering quality of the initial population and the starting point of global optimization.

[0008] (2) In the multi-physics state-driven adaptive evolution algorithm constructed in this invention, multi-physics state-driven adaptive weights are adopted, breaking the limitations of traditional pure time-driven (such as linear or cosine decreasing) convergence parameters. By real-time monitoring of the multi-physics comprehensive state margin of the motor during the optimization process, when the power is sufficient, the algorithm maintains a high weight to squeeze the noise reduction space of high-frequency force waves to the extreme; once the noise reduction variation causes the average torque to fall below the safety threshold, the weight decreases sharply in an exponential nonlinear manner, forcing the algorithm to immediately switch to a local correction mode to restore power. This makes the convergence trajectory of the algorithm completely controlled by the real physical game between noise reduction and power preservation inside the motor, realizing high-precision optimization decision-making.

[0009] (3) In the multi-physics state-driven adaptive evolution algorithm constructed in this invention, frequency domain feature feedback calibration mutation is adopted to replace conventional (such as Cauchy or Gaussian) mutation that relies on probability-based blind trial and error. By performing a Fast Fourier Transform (FFT) on the time-series torque signal output by electromagnetic simulation, the core spatial harmonic amplitudes that cause low-frequency vibrations and high-frequency noise are accurately extracted, and the variable asynchronous length is dynamically controlled by this real physical characteristic quantity. When the high-order harmonic distortion is severe, a large step size perturbation is automatically applied to forcibly destroy the poor pole-slot fit dimensions; when the harmonic suppression meets the standard, the variable asynchronous length is automatically decayed. This mutation based on physical state feedback greatly enhances the targeting of optimization, completely eliminates the technical defect of conventional blind mutation destroying the converged optimal solution in the later stage of iteration, significantly improves the dynamic balance between global exploration and local development of the algorithm, and solves the technical bottleneck of traditional algorithms being prone to getting trapped in local optima. (4) This invention uses a multi-physics field state-driven adaptive evolution algorithm to jointly optimize the position parameters and dimensions of the V-type rotor magnet, taking into account both global jumps and local searches. An average torque constraint penalty term is introduced into the objective function, which can reduce the peak-to-peak value of the cogging torque and the rated torque pulsation value under no-load conditions while ensuring the rated power output (average output torque) of the motor, and ensure the operating efficiency of the motor. It suppresses electromagnetic noise from the physical source and improves the optimization efficiency and accuracy of the motor. Attached Figure Description

[0010] Figure 1 A flowchart illustrating the optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of an exemplary permanent magnet synchronous motor; Figure 3 A comparison chart showing the cogging torque of the motors under no-load conditions before and after optimization; Figure 4 A comparison chart showing the output torque of the motor before and after load optimization; Figure 5 A comparison diagram showing the air gap magnetic flux density of the motor before and after optimization. Detailed Implementation

[0011] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0012] Please see Figure 1 This invention provides an optimization method for permanent magnet synchronous motors in new energy vehicles based on an adaptive evolution algorithm, comprising steps S1 to S4: Step S1: Establish a two-dimensional simulation model of the permanent magnet synchronous motor for new energy vehicles.

[0013] Please see Figure 2 In this embodiment, the permanent magnet synchronous motor to be optimized includes a stator 4 and a rotor 7. The stator 4 includes a stator winding 2, a stator auxiliary slot 3 and a stator yoke 1. The rotor 7 is provided with a permanent magnet 6, a first rivet hole 8, a second rivet hole 9 and a magnetic isolation bridge 5.

[0014] Then, the various dimensional parameters of the motor are determined. In this embodiment, the outer diameter of the stator core is 246mm, the inner diameter of the stator core is 166mm, the outer diameter of the rotor core is 164.4mm, the inner diameter of the rotor core is 63mm, the number of auxiliary slots in the stator is 48, and the number of poles in the rotor is 8. The diameter of the first rivet hole is 1mm, and the diameter of the second rivet hole is 3mm.

[0015] Then, finite element modeling of the motor is performed to determine its material properties, operating data, design variables, and optimization objectives. In this embodiment, ANSYS Maxwell software is used to establish a two-dimensional simulation model of the permanent magnet synchronous motor. The stator and rotor materials are M19_29G, the permanent magnet material is NdFe35N, and the stator windings are made of copper wire. The material density of silicon steel sheet M19_29G is 7650 kg / m³. 3 The density of the permanent magnet is 7400 kg / m³. 3 The density of copper wire is 8933 kg / m 3 After assigning material properties to each component, further simulations were conducted to obtain various electromagnetic performance parameters of the permanent magnet synchronous motor. The design variables are the geometric parameters of the rotor V-shaped magnets, specifically including: the included angle of the inner side of the V-shaped magnets, the position of the magnet pole arc end, the magnet thickness, and the magnet width. The optimization objective is to minimize cogging torque, torque ripple, and air gap magnetic flux density while meeting the average output torque and efficiency requirements under rated operating conditions, in order to achieve noise reduction.

[0016] Step S2: Construct a multi-physics state-driven adaptive evolution algorithm. The multi-physics state-driven adaptive evolution algorithm is based on the pangolin optimization algorithm, introduces a heuristic initialization population strategy based on physical sensitivity, and adopts multi-physics state-driven adaptive weights and frequency domain feature feedback calibration mutation.

[0017] Among them, the heuristic initialization population strategy based on physical sensitivity first establishes reasonable physical boundaries for the position and size parameters of the V-shaped magnet, then extracts the local sensitivity of each size parameter to motor torque pulsation and air gap magnetic flux density harmonics through prior finite element calculation or analytical method, and constructs an adaptive Gaussian sampling distribution by combining the harmonic sensitivity weighting matrix, and finally generates the initial population matrix.

[0018] This strategy specifically includes: establishing upper and lower bounds for the position and size parameters of the rotor magnets, and establishing a local sensitivity model for each optimized variable to motor torque ripple and air gap magnetic flux density harmonics. The local sensitivity within the local sensitivity model is calculated using the following formula:

[0019] in, For the first Changes in design variables The amount of change in torque ripple Local sensitivity; Then, based on the sensitivity, the standard deviation of individual locations is constructed, thereby generating the population individual locations within the actual physical size range. The adaptive sampling standard deviation and individual physical locations are calculated using the following formulas:

[0020]

[0021] in, For the first Adaptive sampling standard deviation of dimensional design variables; This is the scaling factor; To prevent extremely small constants with a denominator of zero; and The first Upper and lower bounds of dimensional design variables; Indicates the first The individual in the first The actual physical dimensions of the design variables; For the first The basic empirical dimensions of design variables; This indicates that the mean is 0 and the standard deviation is 0. It follows a normal distribution.

[0022] This physical sensitivity-based heuristic initialization population strategy enables the initial population to simultaneously avoid high torque pulsation regions and high harmonic distortion regions, significantly improving the engineering quality of the initial population.

[0023] The algorithm also employs a multi-physics-field state-driven adaptive weighting mechanism, which not only dynamically adjusts the iterative weights based on the current motor average torque compliance status, but also forms a state feedback quantity based on efficiency, torque ripple, loss, and dominant harmonic amplitude.

[0024] Specifically, the adaptive weights driven by multiphysics state are obtained through the following steps: First, calculate the multiphysics integrated state margin, expressed as:

[0025] in, For the first The multiphysics integrated state margin corresponding to the optimal individual in the next iteration; , , , , These are the state weight coefficients; , , , and The first The average output torque, efficiency, torque ripple, loss, and dominant harmonic amplitude of the optimal individual at each iteration; This represents the rated average output torque requirement. The efficiency constraint threshold; , and These are reference values ​​for torque ripple, loss, and harmonic amplitude, respectively. Then, based on the logistic function and combined with the multiphysics integrated state margin, the adaptive weights are calculated, as expressed in the following expression:

[0026] in, For the first Adaptive weights for each iteration; and These are the upper and lower bounds of the adaptive weights, respectively; This is the steepness adjustment coefficient for the logistic function.

[0027] The aforementioned adaptive weighting mechanism ensures that the algorithm's convergence behavior is simultaneously constrained by dynamics, efficiency, losses, and electromagnetic noise, rather than being controlled solely by the number of iterations or a single torque index.

[0028] In this embodiment, the behavior update of the multiphysics state-driven adaptive evolution algorithm is divided into two modes: exploration and development. When random number At this time, the exploration mode is executed:

[0029] When random number At that time, execute the development mode:

[0030] in, For the first During the nth iteration The position vectors of each individual; For the first During the nth iteration The position vectors of each individual; For the first The globally optimal individual position vector at the next iteration; and These are the upper and lower bound vectors of the design variables, respectively; and A random number within the range (0,1); Indicates in Random numbers within a range To optimize the dimensionality of the variables.

[0031] The algorithm employs a frequency domain feature feedback calibration mutation mechanism to generate mutation perturbations. Specifically, the frequency domain feature feedback calibration mutation includes: extracting the torque timing signal and air gap magnetic flux density signal from the electromagnetic simulation output, and performing Fast Fourier Transform on both signals to extract the amplitude of the dominant spatial harmonics. The mutation direction and mutation duration are jointly controlled based on the harmonic amplitude and the design variable harmonic sensitivity. The formula for calculating the position after mutation is as follows:

[0032]

[0033]

[0034] in, For the first In the nth iteration, the mutated individual is at the... The variable values ​​of the design variables; For the first In the nth iteration, the globally optimal individual is at the ... The variable values ​​of the design variables; Indicates the first During the nth iteration Frequency-domain oriented variable asynchronous length of the design variable; For the first Adaptive mutation strength at the next iteration; For the first During the nth iteration The harmonic sensitivity weighting of the design variables; and The first Upper and lower bounds of dimensional design variables; Indicates in A random number within a given range; and These are the lower and upper limits of the variation intensity, respectively; The attenuation coefficient; This represents the maximum number of iterations.

[0035] The aforementioned frequency domain feature feedback calibration variation mechanism deeply binds the variation intensity to the high-frequency harmonics of the electromagnetic field and the sensitivity of variables, so that the variation direction preferentially acts on the design variables that contribute more significantly to noise.

[0036] If the objective function value calculated at the new position after the perturbation is better than the current global optimum, then the original optimal individual is replaced, which greatly increases the probability of the algorithm escaping local optima.

[0037] Step S3: The position and size parameters of the permanent magnet synchronous motor rotor magnet in the two-dimensional simulation model established in step S1 are used as optimization variables. The multi-physics state-driven adaptive evolution algorithm constructed in step S2 is used for iterative optimization, and the objective function values ​​corresponding to different combinations of rotor magnet parameters are recorded.

[0038] The position and size parameters of the rotor magnet specifically include: the included angle of the inner side of the V-shaped magnet, the position of the end of the magnet pole arc, the magnet thickness, and the magnet width.

[0039] In this embodiment, a multiphysics-driven adaptive evolution algorithm is written into MATLAB. A VBS script file is dynamically generated in the MATLAB background, and ANSYS Maxwell is called to perform a two-dimensional electromagnetic transient simulation. The dynamically generated VBS script enables data interaction between MATLAB and ANSYS Maxwell for joint simulation. After each simulation, CSV data such as time-torque, air gap magnetic flux density, and time-loss are automatically exported to a specified path. MATLAB reads this data and performs a multi-objective fitness evaluation. The objective function values ​​include unloaded cogging torque, average output torque, torque ripple value, air gap magnetic flux density harmonic amplitude, efficiency, and loss.

[0040] The objective function value is calculated using a multi-objective integrated fitness function and an average torque constraint penalty term. The multi-objective integrated fitness function simultaneously constrains cogging torque, torque ripple, air gap magnetic flux density harmonics, losses, efficiency, and average output torque. Specifically, when the simulated average output torque is lower than a preset safety threshold, the efficiency is lower than a preset efficiency threshold, or the rotor geometry does not meet manufacturing constraints, a penalty value is applied to the objective function through the average torque constraint penalty term. This guides the algorithm to avoid parameter regions characterized by power attenuation, efficiency reduction, or unmanufacturable structures.

[0041] The multi-objective integrated fitness function and the average torque constraint penalty term satisfy the following equation:

[0042]

[0043] in, For the first During the nth iteration The objective function value for each individual; For the first The load torque pulsation of each individual unit; This is a reference value for torque ripple. For the first The no-load cogging torque of each individual tooth, This is a reference value for cogging torque; For the first The dominant harmonic amplitude of each individual; For the first The loss of each individual; For the first The efficiency of an individual; This is a reference value for efficiency. To constrain penalty items; For the first During the nth iteration The average output torque corresponding to each individual; For the first During the nth iteration The motor efficiency corresponding to each individual unit; For the first During the nth iteration The geometric topological consistency penalty for each individual (specifically, the topological consistency penalty for geometric interference, insufficient thickness of the magnetic bridge, magnet overstepping the boundary, and structural nonmanufacturability). , , , , These are multi-objective weighting coefficients for torque ripple, cogging torque, harmonic amplitude, loss, and efficiency, respectively. , , These are the penalty coefficients corresponding to average torque, efficiency, and geometric topology constraints, respectively.

[0044] Step S4: When the iterative optimization reaches the preset condition, stop the iteration and output the position parameters and size parameters of the rotor magnet corresponding to the recorded optimal objective function value as the optimization result.

[0045] The preset condition can be that the number of algorithm iterations reaches a preset maximum number of iterations. If the improvement rate of the objective function value of the best individual across multiple generations is lower than a preset threshold.

[0046] In this embodiment, to ensure the sampling coverage density of the high-dimensional nonlinear solution space and balance the computation of multiphysics finite element analysis, the initial population size of the algorithm is set to 40 or more, and an adaptive termination criterion based on performance evolution gradient is used instead of a fixed number of iterations. During the optimization process, the algorithm monitors the torque ripple decrease rate of the best individuals for several consecutive generations in real time until a physically convergent global solution is found. The entire optimization process incorporates a geometric topology consistency verification module, which uses a preset boundary constraint mapping function to avoid physical interference with the rotor structure in real time. After iteration, the algorithm outputs the Pareto optimal parameter set after multi-objective game theory, covering the inner angle of the V-shaped magnet, the position of the pole arc end, and the magnet thickness and width parameters, thereby achieving global collaborative optimization of electromagnetic noise suppression and power output performance.

[0047] Figure 3 , Figure 4 , Figure 5 Table 1 shows the simulation comparison results of the motor optimized by the present invention and other algorithms (specifically the traditional elk herd algorithm) with the original motor.

[0048] Table 1

[0049] from Figure 3 As shown in Table 1, the original motor's no-load cogging torque value is 6.15 N·m, the peak value of the no-load cogging torque value of the motor after optimization by other algorithms is 5.83 N·m, while the peak value of the no-load cogging torque value of the motor optimized by this invention is reduced to 5.72 N·m. Figure 4 As shown in Table 1, the peak output torque of the original motor under rated load is 91.6 N·m, while the peak output torque under rated load after optimization by other algorithms is 92.2 N·m. However, the motor optimized by this invention, while ensuring a reduction in cogging torque under no-load conditions, achieves a peak output torque of 94.5 N·m under rated load conditions. Figure 5 It can be seen that the air gap magnetic flux density non-uniformity of the motor optimized by this invention is significantly improved compared to the original motor and motors optimized by other algorithms. Since cogging torque and air gap magnetic flux density non-uniformity are the main sources of electromagnetic noise, this invention can better reduce cogging torque and air gap magnetic flux density non-uniformity, resulting in lower motor noise and smoother operation. Furthermore, this invention increases output torque while reducing cogging torque, ensuring power output.

[0050] In summary, the optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithms according to the above embodiments has the following beneficial effects: (1) In the multiphysics-driven adaptive evolution algorithm constructed in this invention, a heuristic initialization population strategy based on physical sensitivity is introduced to replace the traditional blind chaotic or random mapping. This strategy establishes a local sensitivity model of the rotor V-shaped magnet parameters to electromagnetic torque pulsation and spatial harmonics. Driven by prior knowledge of the electromagnetic field, the initial population is concentrated in the parameter subspace where low-distortion waveforms are likely to be generated for adaptive sampling. This avoids wasting computing power in the degraded parameter region that is prone to generating high-frequency electromagnetic howling from the physical level, and greatly improves the engineering quality of the initial population and the starting point of global optimization.

[0051] (2) In the multi-physics state-driven adaptive evolution algorithm constructed in this invention, multi-physics state-driven adaptive weights are adopted, breaking the limitations of traditional pure time-driven (such as linear or cosine decreasing) convergence parameters. By real-time monitoring of the multi-physics comprehensive state margin of the motor during the optimization process, when the power is sufficient, the algorithm maintains a high weight to squeeze the noise reduction space of high-frequency force waves to the extreme; once the noise reduction variation causes the average torque to fall below the safety threshold, the weight decreases sharply in an exponential nonlinear manner, forcing the algorithm to immediately switch to a local correction mode to restore power. This makes the convergence trajectory of the algorithm completely controlled by the real physical game between noise reduction and power preservation inside the motor, realizing high-precision optimization decision-making.

[0052] (3) In the multi-physics state-driven adaptive evolution algorithm constructed in this invention, frequency domain feature feedback calibration mutation is adopted to replace conventional (such as Cauchy or Gaussian) mutation that relies on probability-based blind trial and error. By performing a Fast Fourier Transform (FFT) on the time-series torque signal output by electromagnetic simulation, the core spatial harmonic amplitudes that cause low-frequency vibrations and high-frequency noise are accurately extracted, and the variable asynchronous length is dynamically controlled by this real physical characteristic quantity. When the high-order harmonic distortion is severe, a large step size perturbation is automatically applied to forcibly destroy the poor pole-slot fit dimensions; when the harmonic suppression meets the standard, the variable asynchronous length is automatically decayed. This mutation based on physical state feedback greatly enhances the targeting of optimization, completely eliminates the technical defect of conventional blind mutation destroying the converged optimal solution in the later stage of iteration, significantly improves the dynamic balance between global exploration and local development of the algorithm, and solves the technical bottleneck of traditional algorithms being prone to getting trapped in local optima. (4) This invention uses a multi-physics field state-driven adaptive evolution algorithm to jointly optimize the position parameters and dimensions of the V-type rotor magnet, taking into account both global jumps and local searches. An average torque constraint penalty term is introduced into the objective function, which can reduce the peak-to-peak value of the cogging torque and the rated torque pulsation value under no-load conditions while ensuring the rated power output (average output torque) of the motor, and ensure the operating efficiency of the motor. It suppresses electromagnetic noise from the physical source and improves the optimization efficiency and accuracy of the motor.

[0053] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A new energy vehicle permanent magnet synchronous motor optimization method based on an adaptive evolution algorithm, characterized in that, include: Step S1: Establish a two-dimensional simulation model of the permanent magnet synchronous motor for new energy vehicles; Step S2: Construct a multi-physics state-driven adaptive evolution algorithm. The multi-physics state-driven adaptive evolution algorithm is based on the pangolin optimization algorithm, introduces a heuristic initialization population strategy based on physical sensitivity, and adopts multi-physics state-driven adaptive weights and frequency domain feature feedback calibration mutation. Step S3: The position and size parameters of the permanent magnet synchronous motor rotor magnet in the two-dimensional simulation model established in step S1 are used as optimization variables. The multi-physics field state-driven adaptive evolution algorithm constructed in step S2 is used for iterative optimization, and the objective function values ​​corresponding to different combinations of rotor magnet parameters are recorded. Step S4: When the iterative optimization reaches the preset condition, stop the iteration and output the position parameters and size parameters of the rotor magnet corresponding to the recorded optimal objective function value as the optimization result.

2. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 1, characterized in that, In step S2, the heuristic initialization population strategy based on physical sensitivity is introduced, specifically including: establishing upper and lower bounds for the position and size parameters of the rotor magnets, and establishing a local sensitivity model for each optimization variable to motor torque ripple and air gap magnetic flux density harmonics. The local sensitivity, adaptive sampling standard deviation, and individual physical position within the local sensitivity model are calculated using the following formulas: in, For the first Changes in design variables The amount of change in torque ripple Local sensitivity; For the first Adaptive sampling standard deviation of dimensional design variables; This is the scaling factor; To prevent extremely small constants with a denominator of zero; and The first Upper and lower bounds of dimensional design variables; Indicates the first The individual in the first The actual physical dimensions of the design variables; For the first The basic empirical dimensions of design variables; This indicates that the mean is 0 and the standard deviation is 0. It follows a normal distribution.

3. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 2, characterized in that, In step S2, the multi-physics state-driven adaptive weights are obtained through the following steps: First, calculate the multiphysics integrated state margin, expressed as: in, For the first The multiphysics integrated state margin corresponding to the optimal individual in the next iteration; , , , , These are the state weight coefficients; , , , and The first The average output torque, efficiency, torque ripple, loss, and dominant harmonic amplitude of the optimal individual at each iteration; This represents the rated average output torque requirement. The threshold is set for efficiency constraints. , and These are reference values ​​for torque ripple, loss, and harmonic amplitude, respectively. Then, based on the logistic function and combined with the multiphysics integrated state margin, the adaptive weights are calculated, as expressed in the following expression: in, For the first Adaptive weights for each iteration; and These are the upper and lower bounds of the adaptive weights, respectively; This is the steepness adjustment coefficient for the logistic function.

4. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 3, characterized in that, In step S2, the behavior update of the multiphysics state-driven adaptive evolution algorithm is divided into two modes: exploration and development. When random number At this time, the exploration mode is executed: When random number At that time, execute the development mode: in, For the first During the nth iteration The position vectors of each individual; For the first During the nth iteration The position vectors of each individual; For the first The globally optimal individual position vector at the next iteration; and These are the upper and lower bound vectors of the design variables, respectively; and A random number within the range (0,1); Indicates in Random numbers within a range To optimize the dimensionality of the variables.

5. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 4, characterized in that, In step S2, the frequency domain feature feedback calibration variation specifically includes: extracting the torque timing signal and air gap magnetic flux density signal output from the electromagnetic simulation, and performing Fast Fourier Transform on the torque timing signal and air gap magnetic flux density signal respectively to extract the amplitude of the dominant spatial harmonics. The variation direction and variation length are jointly controlled based on the harmonic amplitude and the design variable harmonic sensitivity. The formula for calculating the position after variation is: in, For the first In the nth iteration, the mutated individual is at the... The variable values ​​of the design variables; For the first In the nth iteration, the globally optimal individual is at the th... The variable values ​​of the design variables; Indicates the first During the nth iteration Frequency-domain oriented variable asynchronous length of the design variable; For the first Adaptive mutation strength at the next iteration; For the first During the nth iteration The harmonic sensitivity weighting of the design variables; and The first Upper and lower bounds of dimensional design variables; Indicates in A random number within a given range; and These are the lower and upper limits of the variation intensity, respectively; The attenuation coefficient; This represents the maximum number of iterations.

6. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 5, characterized in that, In step S3, the position and size parameters of the rotor magnet specifically include: the included angle of the inner side of the V-shaped magnet, the position of the end of the magnet pole arc, the magnet thickness, and the magnet width.

7. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 6, characterized in that, The objective function values ​​in step S3 include no-load cogging torque, average output torque, torque ripple value, air gap magnetic flux density harmonic amplitude, efficiency, and loss.

8. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 7, characterized in that, The objective function value is calculated using a multi-objective integrated fitness function and an average torque constraint penalty term. The multi-objective integrated fitness function simultaneously constrains cogging torque, torque ripple, air gap magnetic flux density harmonics, losses, efficiency, and average output torque. The multi-objective integrated fitness function and the average torque constraint penalty term satisfy the following equation: in, For the first During the nth iteration The objective function value for each individual; For the first The load torque pulsation of each individual unit; This is a reference value for torque ripple. For the first The no-load cogging torque of each individual tooth, This is a reference value for cogging torque; For the first The dominant harmonic amplitude of each individual; For the first The loss of each individual; For the first The efficiency of an individual; This is a reference value for efficiency. To constrain penalty items; For the first During the nth iteration The average output torque corresponding to each individual; For the first During the nth iteration The motor efficiency corresponding to each individual unit; For the first During the nth iteration Geometric topological consistency penalty for each individual; , , , , These are multi-objective weighting coefficients for torque ripple, cogging torque, harmonic amplitude, loss, and efficiency, respectively. , , These are the penalty coefficients corresponding to average torque, efficiency, and geometric topology constraints, respectively.

9. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 1, characterized in that, In step S1, a two-dimensional simulation model of the permanent magnet synchronous motor is established using ANSYS Maxwell software; in step S3, the multiphysics state-driven adaptive evolution algorithm is run using MATLAB, and the joint simulation data interaction between MATLAB and ANSYS Maxwell is realized by dynamically generating VBS scripts.

10. The optimization method for permanent magnet synchronous motors in new energy vehicles based on adaptive evolution algorithm according to claim 1, characterized in that, The preset condition in step S4 is that the number of algorithm iterations reaches a preset maximum number of iterations. If the improvement rate of the objective function value of the best individual across multiple generations is lower than a preset threshold.

Images