Multi-objective optimization method for permanent magnet synchronous motor of electric vehicle

Abstract

The invention discloses a multi-objective optimization method for a permanent magnet synchronous motor of an electric vehicle, which relates to the technical field of permanent magnet synchronous motor structure optimization and comprises the following steps of: establishing a finite element model of the permanent magnet synchronous motor of the electric vehicle, and determining design variables, optimization parameters and optimization objectives; generating a comprehensive sensitivity coefficient according to the sensitivity of each design variable to the optimization target; calculating a value range of each design variable according to the generated comprehensive sensitivity coefficient, thereby determining a state space vector, designing a reward function according to the state space vector, and obtaining a reward value in a current state; selecting an action to be performed according to the current state space vector, generating a new state space vector after the action is adopted according to the performed action, and calculating an action value function of the new state space vector after the action is adopted through a Q-learning algorithm; and circularly carrying out action selection, updating the action value function, and when the updated action value function meets a convergence condition, stopping circulation to obtain an optimized design variable.

Description

Technical Field

[0001] The invention relates to the technical field of permanent magnet synchronous motor structure optimization, and in particular to a multi-objective optimization method for a permanent magnet synchronous motor of an electric vehicle. Background Art

[0002] With the rapid development of social economy and science and technology, cars have become a very common means of transportation in the current human society. However, due to global warming, lack of oil resources and other reasons, countries around the world have higher requirements for the automobile industry. New energy vehicles not only reduce the dependence of the car itself on oil, but more importantly, can effectively reduce environmental pollution, becoming the first choice of the current automobile industry. This has a significant driving effect on the development of new energy vehicles.

[0003] With the emergence of the concept of new energy vehicles, the development of the electric vehicle industry has received great attention from all walks of life. Researching and manufacturing clean and environmentally friendly pure electric vehicles will become the main trend of the current development of the automotive industry. Permanent magnet synchronous motors are the preferred drive motors for pure electric vehicles, with advantages such as high efficiency and good weak magnetic speed regulation capability.

[0004] Although permanent magnet synchronous motors have great advantages over other motors in electric vehicles, they also have obvious defects. Permanent magnet synchronous motors mainly rely on the interaction between the magnetic field generated by permanent magnets and the magnetic field generated by stator windings to obtain output torque. The permanent magnets themselves will demagnetize when the temperature exceeds a certain level. At the same time, the strength of the magnetic field generated by permanent magnets is difficult to adjust, which leads to difficulties in weak magnetic speed regulation of permanent magnet motors. There is also cogging torque, etc. These are all issues that designers must focus on when designing permanent magnet synchronous motors. Therefore, further optimization design should be done to address issues such as the maximum temperature, power density and torque pulsation under rated conditions of permanent magnet synchronous motors so that their related performance can reach a better state.

[0005] In the prior art, publication number CN118036447A discloses a multi-objective optimization method for a permanent magnet synchronous motor, which optimizes multiple performance targets of the motor through a multi-objective optimization algorithm and a fuzzy Taguchi method, thereby improving the overall performance of the motor. This method solves the problem that traditional optimization can only optimize a single target; A neural network constructs a permanent magnet synchronous motor proxy model. Compared with analytical models and finite element models, the model is simple to construct and accurate in calculation, and can solve the problems of time-consuming optimization and low efficiency. This invention uses the Pearson correlation coefficient method and the comprehensive sensitivity coefficient method to classify the design variables. This method can solve the problem that the number of design variables is large, resulting in reduced accuracy of the proxy model and difficulty in converging the optimization algorithm. However, in this method, although the fuzzy Taguchi method provides a certain flexibility in multi-objective optimization, it may not be able to fully consider the nonlinear relationship and dynamic changes between the variables when dealing with complex motor design variables, and the rules of the fuzzy system rely on experience, which may lead to uncertainty and subjectivity in the results, thereby affecting the final optimization effect. At the same time, although the Pearson correlation coefficient method and the comprehensive sensitivity coefficient method are used to classify the design variables to simplify the model, in actual applications, the relationship between the design variables may be very complex, which may lead to the inability to capture high-order interactive effects between the variables, resulting in important design factors being ignored, thereby affecting the final optimization result.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a multi-objective optimization method for a permanent magnet synchronous motor of an electric vehicle to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The multi-objective optimization method of permanent magnet synchronous motor for electric vehicles includes the following specific steps:

[0010] Establish a finite element model of a permanent magnet synchronous motor for an electric vehicle, determine design variables, optimization parameters and optimization objectives, wherein the design variables include stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, and the optimization parameters include power density, maximum temperature under rated conditions of the motor and torque pulsation, and the optimization objectives are maximizing power density, minimizing maximum temperature under rated conditions of the motor and minimizing torque pulsation;

[0011] Calculate the sensitivity between each design variable and the optimization parameter, and combine the above sensitivity to obtain the comprehensive sensitivity coefficient corresponding to each design variable;

[0012] Calculate the value range of each design variable according to the generated comprehensive sensitivity coefficient, and randomly generate a state space vector according to the value range of each design variable, wherein the state space vector is a sequence including all design variables, and the value of the design variable in the sequence is a numerical value within the value range of the design variable;

[0013] Calculate the optimization parameter value corresponding to the state space vector, substitute the calculated optimization parameter value into the reward function, and obtain the reward value under the current state;

[0014] According to the current state space vector, select the action to be performed and generate a new state space vector after taking the action, wherein the action includes increasing or decreasing the values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, outer magnetic bridge width and rotor outer diameter in the current state space vector;

[0015] Calculate the reward value corresponding to the new state space vector after taking the action, set the learning factor and decay factor, and calculate the action value function of the new state space vector through the Q-learning algorithm;

[0016] The action selection is performed cyclically to continuously update the action value function and set the convergence conditions. When the updated action value function meets the convergence conditions, the loop is stopped. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable.

[0017] Furthermore, the design variables are calibrated to have a stator outer diameter of D1 and a tooth width of B. s1 , slot width is B s2 , slot height is H s1 , yoke thickness is H s2 , the angle between the permanent magnets is θ, and the width of the external magnetic bridge is H rib , the rotor outer diameter is D2, and the calibration optimization parameter power density is P den , the maximum temperature of the motor under rated working conditions is T max , torque pulsation is T pkavg .

[0018] Furthermore, the sensitivity between each design variable and the optimization parameter is calculated according to the determined design variables, wherein the sensitivity calculation between each design variable and the optimization parameter is based on the formula:

[0019]

[0020] In the formula, x i is the value of the i-th optimization variable, y j Indicates that in x i The jth optimization parameter value under i,j) represents the sensitivity of the i-th optimization variable to the j-th optimization parameter, Δ is the disturbance, and its value is fixed. j (x i +Δ) means that when the values ​​of other optimization variables except the i-th optimization variable remain unchanged, the value of the j-th optimization parameter will be increased by Δ. j (x i -Δ) represents the value of the jth optimization parameter after the value of the i-th optimization variable is reduced by Δ under the condition that the values ​​of other optimization variables except the i-th optimization variable remain unchanged, where i=1,2,…,8, corresponding to 8 design variable parameters, and j=1,2,3, corresponding to 3 optimization parameters respectively.

[0021] Furthermore, a comprehensive sensitivity coefficient is generated based on the sensitivity analysis between each design variable and the optimization parameter, wherein the formula for calculating the comprehensive sensitivity coefficient is:

[0022] S z (x i )=ω1*S(x i,1 )+ω2*S(x i,2 )+ω3*S(x i,3 )

[0023] In the formula, S z (x i ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the optimization parameter, S(x i,1 ) represents the comprehensive sensitivity coefficient of the i-th optimization variable to power density, S(x i,2 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the maximum temperature of the motor under rated working conditions, S(x i,3 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to torque ripple, ω1, ω2 and ω3 are different weights, where ω1, ω2 and ω3 are all greater than 0 and ω1>ω2>ω3.

[0024] Furthermore, the value range of each design variable is calculated according to the generated comprehensive sensitivity coefficient, wherein the logic for calculating the value range of each design variable is as follows: obtain the reference values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, and calibrate them respectively. p1 , B ps1 , B ps2 , H ps1 , H ps2 ,θ p , H prib and D p2 ; According to the reference value and comprehensive sensitivity coefficient of each design variable, the value range of each design variable is calculated. The calculation formula is:

[0025] (1-S z (x1))*D p1 ≤D1≤(1+S z (x1))*D p1

[0026] (1-S z (x2))*B ps1 ≤B s1 ≤(1+S z (x2))*B ps1

[0027] (1-S z (x3))*B ps2 ≤B s2 ≤(1+S z (x3))*B ps2

[0028] (1-S z (x4))*H ps1 ≤H s1 ≤(1+S z (x4))*H ps1

[0029] (1-S z (x5))*H ps2 ≤H s2 ≤(1+S z (x5))*H ps2

[0030] (1-S z (x6))*θ p ≤θ≤(1+S z (x6))*θ p

[0031] (1-S z (x7))*H prib ≤H rib ≤(1+S z (x7))*H prib

[0032] (1-S z (x8))*D p2 ≤D2≤(1+S z (x8))*D p2

[0033] According to the value range of each design variable generated, the state space vector is generated, where the state space vector is Includes 8 design variable data.

[0034] Furthermore, the method for calculating the optimization parameter value is: inputting the value of each design variable in the state space vector into the finite element model for analysis to obtain the optimization parameter value corresponding to the state space vector;

[0035] The calculation formula of the reward value is as follows:

[0036]

[0037] Where R is the reward value, ω4, ω5 and ω + They are power density P den , the maximum temperature T of the motor under rated working conditions max and torque ripple T pkavg , where ω4, ω5 and ω6 are all greater than 0 and ω4>ω5>ω6.

[0038] Furthermore, the action value function of the new state space vector after taking the action is calculated by the Q-learning algorithm, wherein the formula for calculating the action value function of the new state space vector after taking the action is:

[0039]

[0040] In the formula, Represents time t, represented by the state space vector With action α t Determine the action-value function, represents the action value function after the update at time t+1, δ represents the learning factor, γ is the attenuation factor, and R t represents the reward value calculated at time t, represents the new state space vector The corresponding action-value function maximum value.

[0041] Further,

[0042] The convergence condition is to set a termination threshold. When the updated action value function is greater than or equal to the set termination threshold, the loop is stopped. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable. The specific logic is:

[0043] when When , it is determined to continue the cycle optimization and update the action value function;

[0044] when When , it is determined to stop the loop and obtain the optimization design variables corresponding to the current state space vector;

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] Establish a finite element model of the permanent magnet synchronous motor of an electric vehicle, determine the design variables, optimization parameters and optimization objectives, calculate the sensitivity between each design variable and the optimization objective based on the determined design variables, and generate a comprehensive sensitivity coefficient based on the sensitivity analysis between each design variable and the optimization objective; calculate the value range of each design variable based on the generated comprehensive sensitivity coefficient, generate a state space vector based on the value range of each design variable generated, design a reward function based on the state space vector, and obtain the reward value under the current state; select the action to be performed based on the current state space vector, generate a new state space vector after the action is taken based on the action taken, calculate the reward value corresponding to the new state space vector after the action is taken, set the learning factor and the attenuation factor at the same time, and calculate the action value function of the new state space vector after the action is taken through the Q-learning algorithm; loop through the action selection, generate a new state space vector after the action is taken, calculate the reward value corresponding to the new state space vector after the action is taken, update the action value function, set the convergence condition, and stop the loop when the updated action value function meets the convergence condition. At this time, the parameters in the state space vector corresponding to the action value function are the optimized permanent magnet synchronous motor design variables. This method uses a deep reinforcement learning algorithm, which can adjust strategies in real time during the optimization process. In a multi-objective optimization environment, it can dynamically adjust parameters according to the current optimization status and feedback to better adapt to complex design spaces and increase the probability of finding the global optimal solution. At the same time, the method based on deep reinforcement learning can handle the trade-off problems of multiple objectives at the same time. Compared with the traditional fuzzy Taguchi method, this method can effectively balance the conflicts between different objectives through a reward mechanism, so that the optimization results are more in line with actual application needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0049] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0050] Example:

[0051] See also Figure 1 , the present invention provides a technical solution:

[0052] A multi-objective optimization method for a permanent magnet synchronous motor of an electric vehicle, the specific steps comprising:

[0053] Step 1: Establish a finite element model of the permanent magnet synchronous motor of an electric vehicle, determine the design variables, optimization parameters and optimization objectives, the design variables include the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, the optimization parameters include power density, maximum temperature under rated working conditions of the motor and torque pulsation, and the optimization objectives are to maximize the power density, minimize the maximum temperature under rated working conditions of the motor and minimize the torque pulsation;

[0054] The calibrated design variables are stator outer diameter D1 and tooth width B. s1 , slot width is B s2 , slot height is H s1 , yoke thickness is H s2 , the angle between the permanent magnets is θ, and the width of the external magnetic bridge is H rib , the rotor outer diameter is D2, and the calibration optimization parameter power density is P den , the maximum temperature of the motor under rated working conditions is T max , torque pulsation is T pkavg ;

[0055] The optimization objectives include improving the power density P of the motor. den , in order to reduce the size and weight of the motor, where the power density P den The ratio of the maximum power output of the motor to the effective weight of the motor; the maximum temperature T of the motor under rated working conditions max Minimize, maximum temperature T maxIt needs to be controlled within 160℃ to ensure the normal operation of the motor; the torque pulsation T pkavg Minimize to ensure smooth operation of the motor. The torque ripple is defined as:

[0056]

[0057] Where, T fkmax is the maximum value of the motor peak torque, T fkmin is the minimum value of the motor peak torque, T fkavg is the average value of the motor peak torque;

[0058] Power density P den And the maximum temperature T under rated conditions max Mainly affected by stator tooth width B s1 , Groove width B s2 , Slot height H s1 Yoke thickness H s2 The influence of the motor temperature rise when running at rated state depends on the motor loss, among which the winding copper loss accounts for a relatively high proportion, and the stator tooth width B s1 , Groove width B s2 , Slot height H s1 Yoke thickness H s2 The size of will have a decisive influence on the area of ​​the wire in the motor stator slot, that is, it affects the stator resistance value, thus affecting the winding copper loss. The strength of the magnetic saturation of the motor stator teeth and yoke has a great impact on the power density P den , so the stator tooth width B s1 and the stator yoke thickness H s2 When the power density P increases den will also increase;

[0059] At the same time, the stator outer diameter of the motor is D1 and the rotor outer diameter is D2 for the maximum temperature T under rated working conditions. max The stator outer diameter D1 and the rotor outer diameter D2 will also have a great impact. The distance between the winding and the motor cooling water channel. The closer the winding is to the water channel, the lower the temperature rise of the winding. The part with the highest temperature in the motor is often the winding. Therefore, the stator outer diameter D1 and the rotor outer diameter D2 will affect the maximum temperature T under rated working conditions. max Make an impact;

[0060] The torque ripple of the motor is T pkavg Mainly affected by the permanent magnet angle θ and the width H of the external magnetic bridge rib This is because the position of the permanent magnet and the width of the magnetic isolation bridge have a decisive influence on the harmonic distribution of the air gap magnetic field.

[0061] Step 2: Calculate the sensitivity between each design variable and the optimization parameter, and combine the above sensitivity to obtain the comprehensive sensitivity coefficient corresponding to each design variable;

[0062] The sensitivity between each design variable and the optimization parameter is calculated according to the determined design variables, wherein the sensitivity calculation between each design variable and the optimization parameter is based on the formula:

[0063]

[0064] In the formula, x i is the value of the i-th optimization variable, y j Indicates that in x i The jth optimization parameter value under i,j ) represents the sensitivity of the i-th optimization variable to the j-th optimization parameter, Δ is the disturbance, and its value is fixed. It is set by the staff according to the actual situation and is not limited here. j (x i +Δ) means that when the values ​​of other optimization variables except the i-th optimization variable remain unchanged, the value of the j-th optimization parameter will be increased by Δ. j (x i -Δ) represents the value of the jth optimization parameter after the value of the i-th optimization variable is reduced by Δ under the condition that the values ​​of other optimization variables except the i-th optimization variable remain unchanged, where i=1,2,…,8, corresponding to 8 design variable parameters, and j=1,2,3, corresponding to 3 optimization parameters respectively.

[0065] Among them, S(x i,j ) value, the greater the impact on the optimization parameter. A positive sensitivity index indicates that the optimization parameter is positively correlated with the optimization variable, and a negative sensitivity index indicates that the optimization parameter is negatively correlated with the optimization variable. The jth optimization target value y in the sensitivity calculation j The electromagnetic-thermal coupling calculation model is used for simulation calculation.

[0066] The comprehensive sensitivity coefficient is generated based on the sensitivity analysis between each design variable and the optimization parameter, where the formula for calculating the comprehensive sensitivity coefficient is:

[0067] S z (x i )=ω1*S(x i,1 )+ω2*S(x i,2 )+ω3*S(x i,3 )

[0068] In the formula, S z (x i ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the optimization parameter, S(x i,1) represents the comprehensive sensitivity coefficient of the i-th optimization variable to power density, S(x i,2 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the maximum temperature of the motor under rated working conditions, S(x i,3 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to torque pulsation, ω1, ω2 and ω3 are different weights, among which ω1, ω2 and ω3 are all greater than 0; since the power density has the greatest influence on the motor performance optimization, and the maximum temperature has a greater influence on the motor performance than the torque pulsation, ω1>ω2>ω3.

[0069] Step 3: Calculate the value range of each design variable according to the generated comprehensive sensitivity coefficient, and randomly generate a state space vector according to the value range of each design variable. The state space vector is a sequence including all design variables, and the value of the design variable in the sequence is a value within the value range of the design variable;

[0070] The value range of each design variable is calculated according to the generated comprehensive sensitivity coefficient. The logic for calculating the value range of each design variable is as follows: obtain the reference values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, and calibrate them respectively. p1 , B ps1 , B ps2 , H ps1 , H ps2 ,θ p , H prib and D p2 ; According to the reference value and comprehensive sensitivity coefficient of each design variable, the value range of each design variable is calculated. The calculation formula is:

[0071] (1-S z (x1))*D p1 ≤D1≤(1+S z (x1))*D p1

[0072] (1-S z (x2))*B ps1 ≤B s1 ≤(1+S z (x2))*B ps1

[0073] (1-S z (x3))*B ps2 ≤B s2 ≤(1+S z (x3))*B ps2

[0074] (1-S z (x4))*H ps1≤H s1 ≤(1+S z (x4))*H ps1

[0075] (1-S z (x5))*H ps2 ≤H s2 ≤(1+S z (x5))*H ps2

[0076] (1-S z (x6))*θ p ≤θ≤(1+S z (x6))*θ p

[0077] (1-S z (x7))*H prib ≤H rib ≤(1+S z (x7))*H prib

[0078] (1-S z (x8))*D p2 ≤D2≤(1+S z (x8))*D p2

[0079] According to the value range of each design variable generated, the state space vector is generated, where the state space vector is Includes 8 design variable data.

[0080] The value range of each design variable is determined by the comprehensive sensitivity coefficient, wherein the larger the comprehensive sensitivity coefficient of the design variable is, the larger the value range determined by the formula is, and vice versa, the smaller the comprehensive sensitivity coefficient of the design variable is, the smaller the value range determined by the formula is. Determining the value range of each design variable according to the comprehensive sensitivity coefficient can make the design variables with a large influence on the optimization parameters have more values, improve the accuracy of obtaining the optimal parameters, reduce the value range of the design variables with a small influence on the optimization parameters, reduce the calculation burden, and accelerate the convergence speed.

[0081] Step 4: Calculate the optimization parameter value corresponding to the state space vector, substitute the calculated optimization parameter value into the reward function, and obtain the reward value under the current state;

[0082] The optimization target value corresponding to the state space vector parameter is calculated by using an electromagnetic-thermal coupling calculation model for simulation calculation;

[0083] The method for calculating the optimization parameter value is as follows: inputting the values ​​of each design variable in the state space vector into the finite element model for analysis to obtain the optimization parameter value corresponding to the state space vector;

[0084] The calculation formula of the reward value is as follows:

[0085]

[0086] Where R is the reward value, ω4, ω5 and ω6 are the power density P den , the maximum temperature T of the motor under rated working conditions max and torque ripple T pkavg , where ω4, ω5 and ω6 are all greater than 0. Since the power density has the greatest impact on the motor performance optimization and the maximum temperature has a greater impact on the motor performance than the torque ripple, ω4>ω5>ω6.

[0087] For the power density P den , using a logarithmic function helps balance the reward difference between high power density and low power density. In this way, even designs with lower power density can get appropriate rewards, while designs with higher power density can get higher rewards. To avoid the domain problem of the logarithm, the reward value R is proportional to the power density P den is proportional to;

[0088] For the maximum temperature T of the motor under rated working conditions max The square root form is to punish excessive temperature rise, taking into account the nonlinear effect of temperature rise on motor performance. The square root form can reduce the penalty for excessive temperature, so that the model can tolerate a certain degree of temperature during the optimization process without being too penalized. The reward value R is related to the maximum temperature T under the rated working condition of the motor. max Inversely proportional;

[0089] Torque ripple T pkavg The maximum temperature T of the motor under rated working conditions mak Similarly, no further detailed introduction is given.

[0090] Step 5: According to the current state space vector, select the action to be performed and generate a new state space vector after taking the action, wherein the action includes increasing or decreasing the values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter in the current state space vector;

[0091] The method for selecting the action to be performed is to select the action of the current state space vector according to the ε-greedy strategy. The specific steps include determining the initial value of ε to increase the utilization ratio during the training process. Generate a random number b in the range [0,1]; if b < ε, select a random action, randomly selected from the possible action set, otherwise, select the optimal action under the current state space vector, that is, the action corresponding to the maximum action value function. The action value function of the optimal action is usually obtained through a deep Q network (DQN) or other policy networks.

[0092] Step 6: Calculate the reward value corresponding to the new state space vector after taking the action, set the learning factor and decay factor, and calculate the action value function of the new state space vector through the Q-learning algorithm;

[0093] It is an existing technology to calculate the action value function of the new state space vector after taking an action by using the Q-learning algorithm, wherein the formula for calculating the action value function of the new state space vector after taking an action is:

[0094]

[0095] In the formula, Represents time t, represented by the state space vector With action α t Determine the action-value function, represents the action value function after the update at time t+1, δ represents the learning factor, γ is the attenuation factor, which is set by the staff according to the actual situation, R t represents the reward value calculated at time t, represents the new state space vector The corresponding action-value function maximum value.

[0096] The action value function, or Q value, represents the transition from state To start, select action α t , and the cumulative reward that can be obtained after following the strategy; the initialization is required to start the optimization, and the initialization state space vector According to the value range of the design variables, the Q value table is initialized: the Q value of all state-action pairs is initialized to 0, and the Q value is the action value function; the hyper parameters are set: learning factor δ=0.1, attenuation factor γ=0.9.

[0097] Step 7: Loop through the action selection to continuously update the action value function and set the convergence conditions. When the updated action value function meets the convergence conditions, stop the loop. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable.

[0098] The convergence condition is to set a termination threshold. When the updated action value function is greater than or equal to the set termination threshold, the loop is stopped. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable. The specific logic is:

[0099] when When , it is determined to continue the cycle optimization and update the action value function;

[0100] when When , it is determined to stop the loop and obtain the optimization design variables corresponding to the current state space vector;

[0101] Among them, when When it is less than the set termination threshold, it indicates that the parameters in the current state space vector have not reached the expected optimization goal, and it is necessary to continue the cycle optimization to update the action value function. When it is greater than or equal to the set termination threshold, the loop stops. At this time, the parameters in the state space vector are the optimized permanent magnet synchronous motor design variables, and the design variable values ​​of the state space vector at this time are selected as the optimal solution.

[0102] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0103] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A multi-objective optimization method for permanent magnet synchronous motor of electric vehicle, characterized in that: The specific steps include: Establish a finite element model of a permanent magnet synchronous motor for an electric vehicle, determine design variables, optimization parameters and optimization objectives, wherein the design variables include stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, and the optimization parameters include power density, maximum temperature under rated conditions of the motor and torque pulsation, and the optimization objectives are maximizing power density, minimizing maximum temperature under rated conditions of the motor and minimizing torque pulsation; Calculate the sensitivity between each design variable and the optimization parameter, and combine the above sensitivity to obtain the comprehensive sensitivity coefficient corresponding to each design variable; Calculate the value range of each design variable according to the generated comprehensive sensitivity coefficient, and randomly generate a state space vector according to the value range of each design variable, wherein the state space vector is a sequence including all design variables, and the value of the design variable in the sequence is a numerical value within the value range of the design variable; Calculate the optimization parameter value corresponding to the state space vector, substitute the calculated optimization parameter value into the reward function, and obtain the reward value under the current state; According to the current state space vector, select the action to be performed and generate a new state space vector after taking the action, wherein the action includes increasing or decreasing the values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, outer magnetic bridge width and rotor outer diameter in the current state space vector; Calculate the reward value corresponding to the new state space vector after taking the action, set the learning factor and decay factor, and calculate the action value function of the new state space vector through the Q-learning algorithm; The action selection is performed cyclically to continuously update the action value function and set the convergence conditions. When the updated action value function meets the convergence conditions, the loop is stopped. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable.

2. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 1, characterized in that: The calibrated design variables are stator outer diameter D1 and tooth width B. s1 , slot width is B s2 , slot height is H s1 , yoke thickness is H s2 , the angle between the permanent magnets is θ, and the width of the external magnetic bridge is H rib , the rotor outer diameter is D2, and the calibration optimization parameter power density is P den , the maximum temperature of the motor under rated working conditions is T max , torque pulsation is T pkavg .

3. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 2, characterized in that: The sensitivity between each design variable and the optimization parameter is calculated according to the determined design variables, wherein the sensitivity calculation between each design variable and the optimization parameter is based on the formula: In the formula, x i is the value of the i-th optimization variable, y j Indicates that in x i The jth optimization parameter value under i，j ) represents the sensitivity of the i-th optimization variable to the j-th optimization parameter, Δ is the disturbance, and its value is fixed. j (x i +Δ) means that when the values ​​of other optimization variables except the i-th optimization variable remain unchanged, the value of the j-th optimization parameter will be increased by Δ. j (x i -Δ) represents the value of the jth optimization parameter after the value of the ith optimization variable is reduced by Δ under the condition that the values ​​of other optimization variables except the ith optimization variable remain unchanged, where i=1, 2, ..., 8, corresponding to 8 design variable parameters, and j=1, 2, 3, corresponding to 3 optimization parameters respectively.

4. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 3 is characterized in that: The comprehensive sensitivity coefficient is generated based on the sensitivity analysis between each design variable and the optimization parameter, where the formula for calculating the comprehensive sensitivity coefficient is: S z (x i )=ω1*S(x i，1 )+ω2*S(x i，2 )+ω3*S(x i，3 ) In the formula, S z (x i ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the optimization parameter, S(x i，1 ) represents the comprehensive sensitivity coefficient of the i-th optimization variable to power density, S(x i，2 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to the maximum temperature of the motor under rated working conditions, S(x i，3 ) is the comprehensive sensitivity coefficient of the i-th optimization variable to torque pulsation, ω1, ω2 and ω3 are different weights, where ω1, ω2 and ω3 are all greater than 0 and ω1>ω2>ω3.

5. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 4 is characterized in that: The value range of each design variable is calculated according to the generated comprehensive sensitivity coefficient. The logic for calculating the value range of each design variable is as follows: obtain the reference values ​​of the stator outer diameter, tooth width, slot width, slot height, yoke thickness, permanent magnet angle, external magnetic bridge width and rotor outer diameter, and calibrate them respectively. p1 , B ps1 , B ps2 , H ps1 , H ps2 ,θ p , H prib and D p2 ; According to the reference value and comprehensive sensitivity coefficient of each design variable, the value range of each design variable is calculated. The calculation formula is: (1-S z (x1))*D p1 ≤D1≤(1+S z (x1))*D p1 (1-S z (x2))*B ps1 ≤B s1 ≤(1+S z (x2))*B ps1 (1-S z (x3))*B ps2 ≤B s2 ≤(1+S z (x3))*B ps2 (1-C) z (x4))*H ps1 ≤H s1 ≤(1+S z (x4))*H ps1 (1-S z (x5))*H ps2 ≤H s2 ≤(1+S z (x5))*H ps2 (1-S z (x6))*θ p ≤θ≤(1+S z (x6))*θ p (1-S z (x7))*H prib ≤H rib ≤(1+S z (x7))*H prib (1-S z (x8))*D p2 ≤D2≤(1+S z (x8))*D p2 According to the value range of each design variable generated, the state space vector is generated, where the state space vector is Includes 8 design variable data.

6. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 5, characterized in that: The method for calculating the optimization parameter value is as follows: inputting the values ​​of each design variable in the state space vector into the finite element model for analysis to obtain the optimization parameter value corresponding to the state space vector; The calculation formula of the reward value is as follows: Where R is the reward value, ω4, ω5 and ω6 are the power density P den , the maximum temperature T of the motor under rated working conditions max and torque ripple T pkavg , where ω4, ω5 and ω6 are all greater than 0 and ω4>ω5>ω6.

7. The multi-objective optimization method for permanent magnet synchronous motor of electric vehicle according to claim 1, characterized in that: The convergence condition is to set a termination threshold. When the updated action value function is greater than or equal to the set termination threshold, the loop is stopped. At this time, the value in the state space vector corresponding to the action value function is the optimized permanent magnet synchronous motor design variable. The specific logic is: when When , it is determined to continue the cycle optimization and update the action value function; when When , it is determined to stop the loop and obtain the optimization design variables corresponding to the current state space vector; in Represents the updated action-value function at time t+1.

Images