Layered optimization method of permanent magnet reluctance motor based on improved sensitivity analysis and application

Abstract

The invention discloses a permanent magnet reluctance motor layered optimization method based on improved sensitivity analysis and application, and belongs to the field of synchronous reluctance motor design. According to the method, two sensitiveness analysis methods, namely a finite difference method and a Morse method, which highly depend on trajectory calculation, are combined, a parameter fluctuation rate and a corresponding change rate are redefined, and a plurality of groups of Morse parameter matrixes with single parameter fluctuation in adjacent rows are constructed while the reasonability of model parameters is ensured; the finite difference of adjacent row elements of the parameter matrixes is solved to obtain the local sensitivity of the target to the parameters, and the local sensitivity of all the parameters of the multiple groups of parameter matrixes is comprehensively calculated to obtain the global sensitivity. And based on an improved sensitivity analysis result, performing hierarchical processing on optimization parameters and optimization targets, and performing optimization according to levels.

Description

Technical Field

[0001] The present invention belongs to the technical field of synchronous reluctance motors, and more specifically, relates to a permanent magnet reluctance motor hierarchical optimization method and application based on improved sensitivity analysis. Background Art

[0002] In modern industrial society, motors, as key equipment, play an indispensable role in many fields such as manufacturing, energy production, transportation, and home appliances. my country has a very large number of motors, accounting for more than 60% of the total electricity consumption in society, which fully demonstrates the key position of motors in energy consumption.

[0003] Rare earth permanent magnet synchronous motors have been widely used in various advanced industrial equipment and energy-saving products due to their outstanding advantages such as high efficiency and high power density. However, with the continuous expansion of the motor application market, rare earth, as a key raw material for manufacturing such motors, has limited resource reserves and its market price continues to fluctuate sharply, which has aggravated the cost control difficulties and market risks of motor manufacturers.

[0004] In order to effectively alleviate the "rare earth anxiety" dilemma, rare earth-free / low rare earth synchronous reluctance motors have gradually entered people's field of vision and have received widespread attention from all walks of life. This type of motor sets a magnetic barrier on the rotor to make the rotor's direct and alternating axes anisotropic, and then generates magnetic pull to form reluctance torque through the different reluctances of each part of the rotor. This type of motor can achieve higher energy conversion efficiency while using less rare earth materials (called permanent magnet assisted synchronous reluctance motor) or even no rare earth materials (called synchronous reluctance motor), which reduces the dependence on rare earth materials to a certain extent and provides new ideas and directions for the sustainable development of the motor industry.

[0005] However, this type of motor also faces many challenges in the design process. Its rotor adopts a multi-layer magnetic barrier structure. This complex structural design makes the rotor design parameters numerous. When performing traditional global optimization on these design parameters, due to the complex correlation and influence between the parameters, the optimization process usually consumes a lot of time and computing resources, which greatly prolongs the design cycle. Moreover, due to the large search space and the weak convergence ability of the optimization algorithm, it is very easy to fall into the local optimal solution in the process of finding the optimal solution, and it is difficult to truly achieve the global optimal design goal. Ultimately, the obtained motor design solution is not satisfactory in performance, and the potential performance of the motor cannot be fully utilized, which limits its competitiveness and application scope in the market. Therefore, how to break through these technical bottlenecks and realize the efficient optimization of the design parameters of the rare earth-free / low-rare earth synchronous reluctance motor has become one of the key issues to be solved in the current field of synchronous reluctance motor technology. Summary of the invention

[0006] In view of the defects of the existing technology and the improvement requirements, the purpose of the present invention is to provide a hierarchical optimization method and application for a permanent magnet reluctance motor based on improved sensitivity analysis, so as to enhance the convergence of optimization, reduce the optimization time consumption and improve the optimization effect, and improve and perfect the existing technical solutions.

[0007] To achieve the above object, the present invention provides a hierarchical optimization method for a permanent magnet reluctance motor based on improved sensitivity analysis, including:

[0008] Determine the optimization parameters, optimization objectives and corresponding constraint conditions according to the performance requirements of the permanent magnet reluctance motor;

[0009] Set k volatility rates, construct corresponding k parameter matrices P. Each row of the parameter matrix P represents a set of optimization parameters, and each set of optimization parameters determines a unique motor model. Each column represents an optimization parameter. Perform motor performance simulation on each row to obtain n + 1 sets of simulation results. Perform finite difference calculation on the simulation results to obtain the local sensitivity of the optimization objective to each optimization parameter under each volatility rate. Calculate the weighted average of the local sensitivities under each volatility rate to obtain the global sensitivity of the optimization objective to all optimization parameters;

[0010] According to the global sensitivity of the optimization objective to all optimization parameters, layer by layer in the global sensitivity from the largest to the smallest degree of influence of the optimization objective on the optimization parameters (when the degrees of influence are the same, select the optimization parameters and optimization objectives according to the design requirements), obtain the optimization parameters and optimization objectives of each layer, where the degree of influence of the optimization objective on the optimization parameters refers to the number of optimization parameters whose global sensitivity of the optimization objective to the optimization parameters is greater than the preset value; Set the constraint conditions of the first layer, and the constraint conditions of the first layer are the initial values of the optimization parameters, and use the optimization objective of the upper layer as the constraint condition of the lower layer to obtain the constraint conditions of each layer correspondingly;

[0011] Select the optimization parameters that meet the performance requirements on the Pareto front of the optimization calculation results of each layer, update them to the initial values of the next layer of optimization, and perform the next layer of optimization until the last layer to obtain the final optimization result.

[0012] For one optimization process, there are n optimization parameters, which are x1, x2, ……, x n , and each optimization parameter has its corresponding value range (constraint condition); there are m optimization objectives, which are F1, F2, ……, F m , and the optimization objectives can also be selectively set with value ranges (constraint conditions) according to actual needs; there is a functional relationship between the optimization objectives and the optimization parameters:

[0013]

[0014] Different from the traditional global optimization method that performs optimization calculations on all n parameters in the same batch to obtain the Pareto front formed by m objective functions, the hierarchical optimization method adopts a batch-by-batch and progressive optimization method.

[0015] All n optimization parameters x1, x2, ……, x n Can be repeatedly divided into several groups (denote the number of groups as p):

[0016]

[0017] Similarly, all m optimization objectives F1, F2, ……, F m Can be repeatedly divided into p groups:

[0018]

[0019] Perform one batch of optimization on a group of optimization parameters, a group of optimization objectives and their corresponding constraint conditions, then a total of p batches of optimization are carried out. Obviously, the number of optimizations required for hierarchical optimization is more than that of global optimization, but through the grouping of parameters and objectives, it reduces the dimension of the variable space and can effectively shorten the time for single-batch optimization.

[0020] The above is the basic hierarchical optimization method relied on by the present invention, and its core idea is to break down a huge and complex problem into several relatively simple small problems.

[0021] For a hierarchical optimization method, it is necessary to divide n optimization parameters and m optimization objectives into p groups, combine the corresponding constraint conditions, and perform optimization in sequence. The basis for grouping is the sensitivity of the optimization objective to the change of the optimization parameter. The objective that has a greater impact on the parameter change is called high sensitivity, and vice versa is called low sensitivity. In order to better optimize the convergence, the combination of the objective with higher sensitivity and the parameter is divided into the same group, and this process is called sensitivity analysis.

[0022] Common sensitivity analysis methods include: direct derivative method, Pearson correlation coefficient method, finite difference method, response surface method, and Morris method. The direct derivative method is simple and straightforward, but it relies on a clear functional relationship; the Pearson correlation coefficient method does not require knowledge of the functional relationship, but it is only highly accurate for linear relationships. The above two methods are not applicable to synchronous reluctance motors, as the iron cores of this type of motor are prone to saturation, resulting in complex non-linear electromagnetic coupling relationships. The response surface method can handle the interactions between multiple parameters, but in essence, it also seeks the functional relationship between parameters, requires a large amount of pre-calculation, and the obtained results depend on the selection of pre-set hyperparameters, with a large degree of subjectivity, affecting the calculation results. The finite difference method and the Morris method are based on the trajectory and trend of the calculation results to obtain the sensitivity of parameters to the target, greatly reducing the subjectivity of the results. However, the finite difference method needs to calculate the difference components one by one, and the accuracy highly depends on the step size, resulting in a large amount of calculation; the Morris method has a low calculation cost, but it overly relies on the trajectory and sampling, resulting in insufficient accuracy.

[0023] To reduce the calculation cost while improving the accuracy of the calculation results and ensuring the objective accuracy of the results, the present invention fully considers the performance characteristics of permanent magnet reluctance motors, combines the finite difference method with the Morris method, complements each other's advantages, and completes the sensitivity analysis of permanent magnet reluctance motors.

[0024] The core formula of the finite difference method is:

[0025]

[0026] In the formula, S i is the sensitivity of the optimization objective to the change of the i-th optimization parameter, F j (x) represents the simulation result of the j + 1-th row, ΔF i (x) represents the change in the optimization objective caused by the change of the i-th optimization parameter, and δ represents the change rate of the optimization parameter. The obtained sensitivity is only used to measure the magnitude of the influence of the optimization objective on a certain optimization parameter and has nothing to do with positive or negative. Therefore, the absolute value is added to the formula. Theoretically, the closer the sensitivity S i is to 0, the smaller the sensitivity of the optimization objective to the change of the i-th optimization parameter. On the contrary, the larger the sensitivity S i , the greater the influence of the i-th optimization parameter on the optimization objective.

[0027] Obviously, when different base values are adopted, even if the change rates of the optimization parameters are the same, the obtained sensitivities are different. Therefore, the Morris analysis method is used to reduce the influence of the selection of the initial model on the sensitivity analysis results. Establish a parameter matrix:

[0028]

[0029] For a permanent magnet reluctance motor, in the parameter matrix P, each row represents a set of optimized parameters of the motor, which consists of n elements in total. Each set of parameters can obtain a unique motor model according to the parametric model. By performing finite element simulation on the motor model, the corresponding motor performance can be obtained. Each column in the matrix represents a parameter. For example, the first column represents the axial length of different motor models. The element in the i-th row and j-th column of the matrix represents the j-th parameter to be optimized of the i-th motor model. It can be observed that there is only one parameter value that changes between adjacent rows of the matrix. Substituting the motor performances corresponding to the elements of adjacent rows into the finite difference calculation formula, the local sensitivity of the optimization objective to the change of this optimized parameter can be obtained.

[0030] In addition, in order to avoid the numerical value exceeding the limit range after the change of the optimized parameter, resulting in an unreasonable motor model, the change rate δ of the optimized parameter is defined as follows:

[0031] X i = x imin +(x imax - x imin )×Δx

[0032]

[0033] In the formula, x imin and x imax represent the minimum value and the maximum value of the i-th optimized parameter respectively. Δx is the volatility rate, and ratios such as 10%, 20%,..., 100% can be taken. The fluctuation range of the entire optimized parameter is divided into multiple segments. x i is the base value of the i-th optimized parameter, and X i is the value of the i-th optimized parameter after fluctuation. In this way, multiple different parameter matrices P can be established according to different volatility rates Δx. Perform finite element simulation on the motor models represented by each row in the parameter matrix, calculate the corresponding results of the optimization objective, and substitute them into the finite difference calculation formula to calculate a volatility rate Δx, that is, the local sensitivity S i,j under one parameter matrix. Take the weighted average of the sensitivity results of these parameter matrices to obtain the global sensitivity results of each motor performance to the change of each parameter.

[0034] According to the above improved sensitivity analysis results, group the targets and parameter combinations with higher sensitivities in the same group, and place the group with the highest parameter sensitivity in the front to complete the grouping of the optimized parameters and the optimization objectives. Then perform optimization in sequence.

[0035] For a permanent magnet reluctance motor, when the number of groups is 2 / 3 / 4, the calculation accuracy and calculation cost can be better balanced. There are many optimization parameters and objectives in the first layer, ensuring that the overall performance of the motor parameters is at a relatively excellent level; in the second layer, some parameters are optimized according to specific performance characteristics to further improve the motor performance level.

[0036] The above is the technical solution conceived by the present invention. Compared with the prior art solutions, it fully considers the performance characteristics of the permanent magnet reluctance motor, improves the optimization process of this type of motor, decomposes the complex problem of multiple parameters and multiple objectives into simple problems with fewer parameters and fewer objectives, enhances the optimization convergence, reduces the optimization time consumption, and improves the optimization effect; in the decomposition process, the two methods complement each other, improving the calculation accuracy, reducing the calculation cost, and enhancing the objective accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic flow chart of the hierarchical optimization method for a permanent magnet reluctance motor based on improved sensitivity analysis provided by the present invention.

[0038] Figure 2 It is a schematic diagram of a single-pole model of a 36-slot 4-pole three-layer U-shaped magnetic barrier permanent magnet assisted synchronous reluctance motor described in an embodiment of the present invention.

[0039] Figure 3 It is a schematic diagram of the optimized parameters of the rotor of the 36-slot 4-pole three-layer U-shaped magnetic barrier permanent magnet assisted synchronous reluctance motor model described in an embodiment of the present invention.

[0040] Figure 4 It is a schematic diagram of the optimized parameters of the stator of the 36-slot 4-pole three-layer U-shaped magnetic barrier permanent magnet assisted synchronous reluctance motor model described in an embodiment of the present invention.

[0041] Figure 5 It is a global sensitivity diagram of the optimization objectives of the permanent magnet assisted synchronous reluctance motor with respect to the optimization parameters described in an embodiment of the present invention.

[0042] Figure 6 It is a Pareto front diagram of the first layer optimization in the hierarchical optimization process described in an embodiment of the present invention.

[0043] Figure 7 It is a Pareto front diagram of the second layer optimization in the hierarchical optimization process described in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0045] The present invention proposes a hierarchical optimization method and application for a permanent magnet reluctance motor based on improved sensitivity analysis. Aiming at the characteristics of this type of motor with many design parameters and complex non-linear electromagnetic coupling relationships, two sensitivity analysis methods highly dependent on calculation trajectories, namely the finite difference method and the Morris method, are combined. The parameter volatility Δx and the corresponding change rate δ are redefined. While ensuring the rationality of the model parameters, a Morris parameter matrix with single-parameter fluctuations in multiple adjacent rows is constructed. The finite difference is taken for the elements in adjacent rows of the parameter matrix to obtain the local sensitivity of the target to the parameters. The global sensitivity is obtained by comprehensively calculating the local sensitivities of all parameters in multiple groups of parameter matrices. Based on the results of the improved sensitivity analysis, the optimization parameters and optimization objectives are hierarchically processed and optimized according to the levels. The hierarchical method has high calculation accuracy and low calculation cost, and is well applicable to the non-linear complex electromagnetic relationship of the permanent magnet reluctance motor; the hierarchical optimization method has strong convergence, short optimization time, and good optimization effect, and is of great significance for improving the performance of this type of motor and its extended applications.

[0046] Specifically, as Figure 1 shown, the present invention provides a hierarchical optimization method for a permanent magnet reluctance motor based on improved sensitivity analysis, including:

[0047] Determine the optimization parameters, optimization objectives, and corresponding constraint conditions according to the performance requirements of the permanent magnet reluctance motor;

[0048] Set k volatilities, construct k corresponding parameter matrices P. Each row of the parameter matrix P represents a group of optimization parameters, and each group of optimization parameters determines a unique motor model. Each column represents an optimization parameter. Perform motor performance simulations on each row to obtain n + 1 groups of simulation results. Perform finite difference calculations on the simulation results to obtain the local sensitivity of the optimization objective to each optimization parameter under each volatility. Take the weighted average of the local sensitivities under each volatility to obtain the global sensitivity of the optimization objective to all optimization parameters;

[0049] According to the global sensitivity of the optimization objective to all optimization parameters, hierarchically classify in the global sensitivity from the largest to the smallest according to the degree of influence of the optimization objective on the optimization parameters (when the degrees of influence are the same, select the optimization parameters and optimization objectives according to the design requirements) to obtain the optimization parameters and optimization objectives of each layer, where the degree of influence of the optimization objective on the optimization parameters refers to the number of optimization parameters for which the global sensitivity of the optimization objective to the optimization parameters is greater than a preset value; set the constraint conditions of the first layer, and the constraint conditions of the first layer are the initial values of the optimization parameters, and use the optimization objective of the previous layer as the constraint condition of the next layer to correspondingly obtain the constraint conditions of each layer;

[0050] Select the optimization parameters that meet the performance requirements on the Pareto front of the optimization calculation results of each layer, update them to the initial values of the next layer of optimization, and perform the next layer of optimization until the last layer to obtain the final optimization results.

[0051] Furthermore, the parameter matrix P is an (n + 1) * n matrix, where n is the number of optimization parameters; only one parameter changes between adjacent rows of the matrix, in the form of:

[0052]

[0053] Among them, x i is the base value of the i-th optimization parameter, and X i is the value after the i-th optimization parameter fluctuates;

[0054] The functional relationship between the volatility and the change rate δ of the optimization parameter is:

[0055] X i = x imin +(x imax -x imin )×Δx k

[0056]

[0057] Among them, x imin and x imax represent the minimum and maximum values of the i-th optimization parameter respectively, Δx k is the volatility, and k is the number of volatilities.

[0058] Furthermore, the formula for the finite difference calculation is as follows:

[0059]

[0060] In the formula, S i is the sensitivity of the optimization objective to the change of the i-th optimization parameter, F j (x) represents the simulation result of the (j + 1)-th row, and ΔF i (x) represents the change in the optimization objective caused by the change of the i-th optimization parameter, and δ represents the change rate of the optimization parameter.

[0061] The present invention also provides a synchronous reluctance motor designed by the above-mentioned hierarchical optimization method of a permanent magnet reluctance motor based on improved sensitivity analysis.

[0062] The present invention also provides a permanent magnet assisted synchronous reluctance motor designed by the above-mentioned hierarchical optimization method of a permanent magnet reluctance motor based on improved sensitivity analysis.

[0063] The present invention also provides an electronic device, including: a computer-readable storage medium and a processor;

[0064] The computer-readable storage medium is used to store executable instructions;

[0065] The processor is configured to read the executable instructions stored in the computer-readable storage medium and execute the above method.

[0066] The present invention also provides a computer-readable storage medium storing computer instructions for causing a processor to execute the above method for hierarchical optimization of a permanent magnet reluctance motor based on improved sensitivity analysis.

[0067] The present invention also provides a computer program product including a computer program or instructions, which, when executed by a processor, implement the above method for hierarchical optimization of a permanent magnet reluctance motor based on improved sensitivity analysis.

[0068] Embodiment

[0069] A 36-slot 4-pole three-layer U-shaped magnetic barrier permanent magnet assisted synchronous reluctance motor, the single-pole model of which is as Figure 2 shown. This motor is applied to industrial occasions, and mainly focuses on the efficiency η, torque ripple T ripple , material cost R cost , rated torque T dq , peak torque T max , rated current density J N .

[0070] Parametric modeling is performed on the three-layer U-shaped magnetic barrier permanent magnet reluctance rotor, and there are a total of 17 rotor design parameters, as Figure 3 shown. In addition, it is also necessary to optimize 4 stator design parameters, namely stator tooth width w t , yoke height h bi , slot opening width w so , and motor split ratio χ (the ratio of the inner diameter of the stator to the outer diameter of the stator), as Figure 4 shown. In addition, the axial length L stk of the motor is added, for a total of 22 design parameters. The design parameters to be optimized and their corresponding fluctuation ranges in the initial scheme are listed in Table 1.

[0071] Table 1 Initial values and fluctuation ranges of the design parameters to be optimized for the motor

[0072]

[0073]

[0074] According to the concept proposed by the present invention, the above 22 design parameters are the optimization parameters x1, x2,..., x 22 , efficiency η, torque ripple T ripple, Material cost R cost , Rated torque T dq , Peak torque T max , Rated current density J N That is, the optimization objectives F1, F2, ……, F6. To simplify the optimization process while ensuring performance requirements, only the efficiency η, torque ripple T ripple , and material cost R cost are optimized. The value ranges of the remaining three objectives are used as constraint conditions, that is, the rated torque T dq is not less than 84 Nm, the peak torque T max is not less than 430 Nm, and the rated current density J N is not higher than 5 A / mm 2 .

[0075] According to the concept proposed in the present invention, the above "22-parameter - 3-objective - 3-constraint" optimization process is hierarchically disassembled. Taking Δx = 10%, 20%, ……, 100%, 10 23×22 parameter matrices are obtained, and 22×10 finite element calculations are performed to obtain the change trajectories of the motor performance (optimization objectives) under parameter fluctuations. Substituting the trajectories into the finite difference calculation formula, the sensitivities of the optimization objectives to the optimization parameters are obtained, as shown in Figure 5 . Among them, S R represents the sensitivity of the design parameter to the torque ripple at the rated point, which is represented by a blue color bar, and S E represents the sensitivity of the design parameter to the efficiency at the rated point, which is represented by a yellow color bar, and S C represents the sensitivity of the design parameter to the motor cost, which is represented by a red color bar. It should be noted that since the values of each sensitivity vary greatly, for the convenience of display, only the range of the ordinate from 0 to 2 is intercepted in the figure. The black dotted line is the set threshold. The color bar above the threshold represents that the performance has a high sensitivity to parameter changes, and below the threshold represents that the performance has a low sensitivity to parameter changes.

[0076] The specific values of each sensitivity are shown in Table 2.

[0077] Table 2 Sensitivities of motor optimization objectives to changes in optimization parameters

[0078]

[0079]

[0080] Except for the axial length of the motor, the design parameters are generally more sensitive to the torque ripple at the rated point, and the stator slot opening width and the rotor magnetic barrier tail angle are particularly significant. This is because small fluctuations in the design parameters will change the harmonic distribution of the stator and rotor magnetomotive forces, thereby having a greater impact on the value of the torque ripple. However, the axial length has a linear relationship with the instantaneous torque and thus has no effect on the torque ripple. For the efficiency of the motor, changes in the design parameters will not cause as large a change as the torque ripple. Therefore, the sensitivity values of the design parameters to the efficiency are generally lower than those to the torque ripple. For the cost of the motor, the split ratio of the motor and the parameters of the stator teeth and yoke affect the slot area and thus the copper cost, and the width of the permanent magnet affects the permanent magnet cost. Therefore, these design parameters are more sensitive to the cost. The above qualitative analysis further verifies the correctness of the improved sensitivity analysis results.

[0081] There are the most design parameters with high sensitivity to the torque ripple at the rated point, followed by the rated point efficiency, and the fewest are the design parameters with high sensitivity to the cost of the motor. Therefore, all design parameters are divided into three layers according to this characteristic.

[0082] First, the design parameters that are highly sensitive to the three optimization objectives of the torque ripple at the rated point, the rated point efficiency, and the cost of the motor are classified into the first layer. The optimization objectives of this layer are set to minimize the torque ripple at the rated point of the motor, maximize the rated point efficiency, and minimize the cost of the motor. In the optimization, the inequality constraint conditions are set as follows: the rated current density of the motor is less than 5 A / mm 2 , the rated torque is greater than 84 Nm, and the peak torque is greater than 430 Nm. During optimization, except for the parameters in the first layer being set as the parameters to be optimized, the remaining design parameters are fixed to the parameter values of the initial scheme. After the first layer optimization is completed, select the scheme with a rated point efficiency not lower than the global optimal scheme and lower costs and torque ripples as the optimal scheme for the first layer optimization.

[0083] Secondly, the design parameters that are highly sensitive to the two optimization objectives of the torque ripple at the rated point and the cost of the motor are classified into the second layer. The optimization objectives of this layer are set to minimize the torque ripple at the rated point of the motor and minimize the cost of the motor. Although the parameters in the second layer have low sensitivity to the rated point efficiency, considering that the sensitivity is not equal to 0, that is, the optimization process of the parameters in the second layer may still affect the rated point efficiency. Therefore, in the inequality constraint conditions, in addition to the three conditions in the first layer optimization, an additional constraint condition needs to be added that the rated point efficiency is not lower than the rated point efficiency of the optimal scheme in the first layer. During optimization, except for the parameters in the second layer being set as the parameters to be optimized, the remaining design parameters are fixed to the parameter values of the optimal scheme in the first layer. After the second layer optimization is completed, select the scheme with a motor cost not lower than the global optimal scheme and lower torque ripples as the optimal scheme for the second layer optimization.

[0084] Finally, the design parameters that are highly sensitive only to the rated-point torque ripple are classified into the third layer, and the optimization objective of this layer is set to minimize the rated-point torque ripple of the motor. Similarly, in the inequality constraint conditions, in addition to the 4 conditions in the second-layer optimization, the constraint condition that the motor cost is not higher than the motor cost of the optimal solution in the second layer needs to be added. During optimization, except for the parameters in the third layer being set as the parameters to be optimized, the remaining design parameters are fixed to the parameter values of the optimal solution in the second layer. After the third-layer optimization is completed, the solution with the lowest rated-point torque ripple is selected as the optimal solution for the third-layer optimization, that is, the hierarchical optimal solution.

[0085] Considering that only the axial length L stk This design parameter is relatively special and has low sensitivity to torque ripple. According to the above-established method, it cannot be classified into any layer. Moreover, the axial length has high sensitivity to the other two target performances. Therefore, it is classified into the first layer and the second layer. After the first-layer optimization is completed, the axial length is still set as the parameter to be optimized.

[0086] According to the above method, the 22 design parameters of the permanent magnet assisted synchronous reluctance motor are divided into 3 layers. The optimized parameters, optimization objectives, basis for scheme selection, and supplementary constraint conditions for each layer are shown in Table 3.

[0087] Table 3 Hierarchical optimization results of the hierarchical optimization method conceived in the present invention

[0088]

[0089]

[0090] First, optimize the parameters of the first layer of hierarchical optimization. Considering that there are 8 parameters in the first layer, the number of evolutionary generations is set to 60, and 80 individuals are generated in each generation. Adding the initial individuals, a total of 4880 motor schemes are generated. The distribution of the motor performances that meet the constraint conditions in the three-dimensional target space and the projections on the three two-dimensional target planes for each generation are as Figure 6 shown. The selected optimal solution for the first layer is marked as a purple hollow circle in Figure 6 The rated-point torque ripple of this solution is 32.97%, the rated-point efficiency is 95.08%, and the motor cost is 982.8 yuan.

[0091] Next, optimize the parameters of the second layer of hierarchical optimization. Considering that there are 11 parameters to be optimized in the second layer, the number of evolutionary generations is set to 60, and 110 individuals are generated in each generation. Adding the initial individuals, a total of 6710 motor schemes are generated. The distribution of the motor performances that meet the constraint conditions in the two-dimensional target space for each generation is as Figure 7 shown. The selected optimal solution for the second layer is in Figure 7It is marked as a green hollow circle. The rated point torque ripple of this scheme is 20.05%, the rated point efficiency is 95.07%, and the motor cost is 964.2 yuan.

[0092] Finally, optimize the parameters of the third layer of hierarchical optimization. Considering that there are only 4 parameters to be optimized in the third layer, set the number of evolutionary generations to 20, generate 40 individuals per generation, and add the initial individuals, a total of 840 motor schemes are generated. Since the optimization goal is only to minimize the torque ripple, the individual distribution during this optimization process is no longer shown in the figure. The optimal scheme of the third layer selected is the hierarchical optimal scheme. The rated point torque ripple of this scheme is 16.46%, the rated point efficiency is 95.05%, and the motor cost is 964.0 yuan. The total number of motor schemes for sensitivity analysis and hierarchical optimization is 12,650.

[0093] The optimized target performances of the globally optimal and hierarchically optimal motor schemes are compared in the following table. In addition, both optimization processes are carried out on the same computer platform, which is a desktop workstation with 16 cores and 128 GB of memory. The finite element simulation time consumed by each motor scheme is about 20 seconds. The optimization time is calculated based on the number of motor schemes generated by the two optimization methods and is also listed in Table 4.

[0094] Table 4 Comparison of the hierarchical optimization scheme with the initial scheme and the traditional global optimization scheme

[0095]

[0096] It can be found that in terms of the optimization effect, when the rated efficiency and motor cost of the hierarchically optimal scheme are basically the same as those of the globally optimal scheme, the rated point torque ripple is reduced by two percentage points. In terms of the optimization time, the hierarchical optimization method only takes 71% of the optimization time of the global optimization method, that is, the hierarchical optimization method obtains better optimization results with less time. The reason is that the hierarchical optimization method decomposes a global optimization problem of "22 parameters - 3 objectives - 3 constraint conditions" into a progressive optimization problem of "8 parameters - 3 objectives - 3 constraint conditions" plus "11 parameters - 2 objectives - 4 constraint conditions" plus "4 parameters - 1 objective - 5 constraint conditions". The number of parameters to be optimized in each layer decreases, so the convergence speed of the algorithm is fast, and the required number of optimization generations can be reduced accordingly. In addition, as the number of parameters to be optimized in each layer decreases, the cross-phenomenon of the motor model caused by the fluctuation of the parameters within the value range also decreases, and the number of effective motor schemes generated during the optimization process increases, improving the optimization efficiency. Since the optimization goal and limiting conditions of each layer are specified according to the sensitivity of the design parameters of this layer to the motor performance, the optimization target is more clearly directed.

[0097] It can be seen that the hierarchical optimization method proposed by the present invention is superior to the traditional global optimization method.

[0098] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A hierarchical optimization method for a permanent magnet reluctance motor based on improved sensitivity analysis, characterized in that Including: Determine the optimization parameters, optimization objectives and corresponding constraint conditions according to the performance requirements of the permanent magnet reluctance motor; Set k volatility rates, construct k corresponding parameter matrices P. Each row of the parameter matrix P represents a set of optimization parameters. Each set of optimization parameters determines a unique motor model. Each column represents an optimization parameter. Perform motor performance simulations on each row to obtain n + 1 sets of simulation results. Perform finite difference calculations on the simulation results to obtain the local sensitivity of the optimization objective to each optimization parameter under each volatility rate. Calculate the weighted average of the local sensitivities under each volatility rate to obtain the global sensitivity of the optimization objective to all optimization parameters; According to the global sensitivity of the optimization objective to all optimization parameters, stratify in the global sensitivity from large to small according to the degree of influence of the optimization objective by the optimization parameters to obtain the optimization parameters and optimization objectives of each layer, where the degree of influence of the optimization objective by the optimization parameters refers to the number of optimization parameters for which the global sensitivity of the optimization objective to the optimization parameter is greater than a preset value; Set the constraint conditions of the first layer, and use the optimization objective of the upper layer as the constraint condition of the lower layer to correspondingly obtain the constraint conditions of each layer; Select the optimization parameters that meet the performance requirements on the Pareto front of the optimization calculation results of each layer, update them to the initial values of the lower layer optimization, and perform the lower layer optimization until the last layer to obtain the final optimization result.

2. The hierarchical optimization method for a permanent magnet reluctance motor based on improved sensitivity analysis according to claim 1, wherein The parameter matrix P is an (n + 1) * n matrix, where n is the number of optimization parameters; only one parameter changes between adjacent rows of the matrix, in the form of: where x i is the base value of the i-th optimization parameter, and X i is the value after the fluctuation of the i-th optimization parameter; The functional relationship between the volatility rate and the optimization parameter change rate δ is: X i = x imin + (x imax - x imin ) × Δx k where x imin and x imax represent the minimum and maximum values of the i-th optimization parameter respectively, Δx k is the volatility, and k is the number of volatilities.

3. The hierarchical optimization method of a permanent magnet reluctance motor based on improved sensitivity analysis according to claim 2, characterized in that, The formula for the finite difference calculation is as follows: where S i is the sensitivity of the optimization objective to the change of the i-th optimization parameter, F j (x) represents the simulation result of the (j + 1)-th row, and ΔF i (x) represents the change in the optimization objective caused by the change of the i-th optimization parameter, and δ represents the change rate of the optimization parameter.

4. A synchronous reluctance motor, characterized in that, Designed by the hierarchical optimization method of the permanent magnet reluctance motor based on improved sensitivity analysis according to any one of claims 1 to 3.

5. A permanent magnet assisted synchronous reluctance motor, characterized in that, Designed by the hierarchical optimization method of the permanent magnet reluctance motor based on improved sensitivity analysis according to any one of claims 1 to 3.

6. An electronic device, characterized in that, Including: A computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the hierarchical optimization method of the permanent magnet reluctance motor based on improved sensitivity analysis according to any one of claims 1 to 3.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the processor to execute the hierarchical optimization method of the permanent magnet reluctance motor based on improved sensitivity analysis according to any one of claims 1 to 3.

8. A computer program product, characterized in that, Including a computer program or instructions, and when the computer program or instructions are processed and executed, the hierarchical optimization method of the permanent magnet reluctance motor based on improved sensitivity analysis according to any one of claims 1 to 3 is implemented.

Images