阅读说明：本技术 一种基于熵权法的永磁同步电机多目标参数优化方法 (Entropy weight method based permanent magnet synchronous motor multi-target parameter optimization method ) 是由 李建贵 柯少兴 陈豪 郝诚 于 2020-05-28 设计创作，主要内容包括：一种基于熵权法的永磁同步电机多目标参数优化方法,包括以下步骤：第一步：建立基于模型补偿自抗扰的永磁同步电机矢量控制Simulink模型；第二步：运用综合评价法对基于模型补偿自抗扰的永磁同步电机矢量控制系统模型进行多目标分层并建立相应的评价指标体系；第三步：针对同一性能下的多指标采取熵权法进行客观分配各指标权值,针对控制系统的超调问题采取罚函数进行无约束处理；第四步：选取优化目标函数,运用基于非支配解排序的NSGA-Ⅱ算法整定与优化模型补偿参数,对基于模型补偿自抗扰的永磁同步电机矢量控制系统仿真模型的响应快速性和平稳性进行多目标优化。本设计不仅采用客观的方法进行多目标参数优化,而且有效提高电机的整定效率。(A permanent magnet synchronous motor multi-objective parameter optimization method based on an entropy weight method comprises the following steps: the first step is as follows: establishing a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection; the second step is that: carrying out multi-target layering on the model of the PMSM vector control system based on model compensation active disturbance rejection by using a comprehensive evaluation method and establishing a corresponding evaluation index system; the third step: an entropy weight method is adopted for multiple indexes under the same performance to objectively distribute each index weight, and a penalty function is adopted for overshoot problem of a control system to carry out unconstrained processing; the fourth step: and selecting an optimization target function, and performing multi-objective optimization on the response rapidity and the stability of the PMSM vector control system simulation model based on model compensation active disturbance rejection by using NSGA-II algorithm setting and optimization model compensation parameters based on non-dominated solution sorting. The design not only adopts an objective method to carry out multi-objective parameter optimization, but also effectively improves the setting efficiency of the motor.)

1. A permanent magnet synchronous motor multi-objective parameter optimization method based on an entropy weight method is characterized in that:

the multi-objective parameter optimization method comprises the following steps:

the first step is as follows: establishing a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection,

the controlled object is a mathematical model of the surface-mounted permanent magnet synchronous motor, and is shown in formula (1):

u in formula (1)_dStator voltage of d-axis, u_qStator voltage of q-axis, i_dStator voltage of d-axis, i_qStator current of q-axis, L_dStator inductance of d-axis, L_qStator inductance of q axis, R stator phase resistance, w_eIs the mechanical angular velocity, psi, of the motor_fIs a mechanical angle flux linkage of the motor, P_nIs the magnetic pole pair number, J is the rotational inertia of the motor, B is the damping coefficient of the motor, T_eIs the electromagnetic torque of the machine, T_LIs the load torque of the motor;

the surface-mounted permanent magnet synchronous motor adopts_d ^*The method comprises the following steps that (1) a double closed-loop vector control strategy is set to 0, a space vector pulse width modulation technology is applied in vector control to carry out variable frequency speed regulation on a motor, a sliding mode disturbance observer is utilized to carry out model compensation on a second-order linear active disturbance rejection controller in a rotating speed loop, and the formula (2) is as follows:

in the formula (2), y is the system rotating speed output w_e，z₁Differential tracking signal of y, z₂Is z₁Of the differential tracking signal, z₃For differential tracking signals of system disturbances, e is w_eβ, β₁Gain corrected for zero order of output error, β₂To output a first order correction gain for the error, β₃Is the second-order correction gain of the output error, u is the second-order linear active disturbance rejection controller rotational speed output of the model compensation,

the current loop in the vector control of the surface-mounted permanent magnet synchronous motor adopts PI control, and in addition, the estimated output of a sliding mode disturbance observer is applied

k in formula (3)_pdProportional gain, K, of d-axis current loop PI controller_pqProportional gain, K, of a q-axis current loop PI controller_idD-axis current loop PI controller integral gain, K_iqCurrent loop PI controller integral gain, i, for q-axis_d ^*Current loop command input value, i, for d-axis_q ^*Is the current loop instruction input value of the q axis, s is a Laplace operator,for the d-axis current loop voltage compensation value,a current loop voltage compensation value for the q axis;

the key point in the feedforward compensation control strategy is to use the estimated value of the sliding mode disturbance observer(i.e., disturbance estimate) is introduced into the compensation control of the current loop and a compensation parameter k is set_cd、k_cqAdjusting, wherein the specific design expression is shown as a formula (4);

k in the formula (4)_cdCurrent loop compensation parameter, k, for d-axis_cqCurrent loop of q axisA compensation parameter;

therefore, a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection is built through the design of a rotating speed loop and a current loop and the related space vector pulse width modulation technology;

the second step is that: carrying out multi-target layering on a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection by using a comprehensive evaluation method and establishing a corresponding evaluation index system;

the establishment of the evaluation index system comprises the steps of taking the control system with the aspects of overshoot, response rapidity, stationarity and accuracy as a target decision layer;

respectively confirming index parameters influencing the objective decision layer: by peak time t_pRising time t_rAnd adjusting the time t_sThe method is characterized by comprising the following steps of (1) taking a response rapidness index as a characteristic, taking a rotating speed steady-state error as an accuracy index as a characteristic, taking a rotating speed volatility tau and a rotating speed residual error e as a stability index as a characteristic, and taking an overshoot as an overshoot performance index as a characteristic; seven performance indexes representing four aspects of the system are used as index layers;

the specific calculation of the index layer is that on the basis of the Simulink model established in the first step, a target rotating speed w is set_r1000r/min, actual output value w of rotating speed when starting under no-load condition_eSeven indices were calculated as follows:

i rapidity index: time of peak t_pRising time t_rAdjusting the time t_sTime of peak t_pTime taken for the first time to reach the maximum output of the rotation speed waveform, rise time t_rAdjusting the time t for the first time to reach the set rotation speed_sAdjusting the time t for the time when the set rotation speed allowable error range is reached for the first time in the steady state process_sThe allowable error range of (2%);

ii accuracy index: the rotating speed steady-state error is the difference value between the rotating speed output value and the rotating speed average value in the steady-state process, and the calculation formula is shown as the formula (5):

＝w_e-w_{e_mean}(5)

iii smoothness index: the rotation speed fluctuation tau is the ratio of the maximum and minimum difference values of the rotation speed in the steady state process, and the calculation formula is shown as the formula (6):

w in the formulae (5) and (6)_eActual output value of the rotational speed at no-load start, w_{e_max}Is the actual output value w of the rotating speed in the steady state process_eMaximum value of, w_{e_min}Is the actual output value w of the rotating speed in the steady state process_eMinimum value of, w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the rotating speed residual e is the difference value between the output value of the rotating speed and the set value of the rotating speed, and the calculation formula is shown as the formula (7):

e＝w_e-w_r(7)

iv, overshooting performance indexes: overshoot, overshoot being the peak time t_pThe calculation formula of the ratio of the maximum deviation value of the lower rotating speed output to the output value in the steady state process is shown as the formula (8):

w in the formulae (7) and (8)_rTarget rotational speed set for the control system (i.e. 1000r/min), w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the third step: regularizing the data matrixes with different performance indexes in the index evaluation system established in the second step, reasonably distributing the weight of each index to multiple indexes with different performances, and preferentially processing the overshoot if the overshoot problem exists in the control system, so that penalty function processing is performed on the overshoot, and entropy weight processing is performed on the multiple indexes with the weight needing to be distributed;

the overshoot is subject to a penalty function, if overshoot exists, the coefficient given by the index is far larger than the coefficients of other indexes, an entropy weight method is adopted to objectively distribute weight values for three indexes (namely, rotational speed volatility tau, rotational speed residual error e and steady-state error) for representing the stability and the accuracy of a control system, and finally the indexes of the rotational speed volatility tau, the rotational speed residual error e, the steady-state error and the overshoot are integrated into an optimized objective function, wherein the calculation formula is shown as formula (9):

g₁(t)＝w₁·τ+w₂·e+w₃·+c· (9)

w in formula (9)₁For objective assignment of the weight, w, of the rotational speed ripple τ₂For objective distribution of weight sum w of rotation speed residual e₃For objective assignment of weight, w, of steady-state error₁、w₂And w₃The sum is equal to 1, c is a penalty factor and is far more than 0, c is 100, g₁(t) is a first optimization objective function;

in the evaluation index system, three indexes in seven indexes are as follows: the rotation speed volatility tau, the rotation speed residual error e and the steady-state error are distributed with weight by adopting an entropy weight method, and the three calculation processes of the entropy weight method are as follows:

the first process is as follows: and (3) data standardization, namely standardizing each index data:

in the formula (10), i is an index data sequence, j is an index number, and x_{i_max}Is the maximum value of j index, x_{i_min}Is the minimum value of the j index;

and a second process: calculating the information entropy of each index:

in the formula (11)Is the proportion of single data in the j target lower index;

and a third process: determining the weight of each index:

in the formula (12), j is 1,2,3, …, and n is the weight of each index;

the peak time t for representing the response rapidity in the evaluation index system_pRising time t_rAdjusting the time t_sThe three indexes are all related to time, so the time minimization principle is adopted for design; and combining a penalty function of the overshoot to emphatically prevent the control system from generating overshoot; therefore, the peak time t_pRising time t_rAdjusting the time t_sAnd the indexes of overshoot are integrated into another optimization objective function, and the calculation formula is shown as the formula (13):

g(t)＝t_p+t_r+t_s+c· (13)

in formula (13), g₂(t) is a second optimization objective function, c is a penalty factor and is far greater than 0, and the value of c is 100;

therefore, the established evaluation index system is quantized to obtain two optimized objective functions: the first is an optimization objective function related to smooth and accurate output of the rotation speed, see equation (9), and the second is an optimization objective function related to time (i.e., response rapidity), see equation (13);

the fourth step: applying the two optimized objective functions obtained in the third step to a compensation parameter k in a Simulink model by using a non-dominated solution sorting-based NSGA-II algorithm_cd、k_cqOptimizing so as to realize the optimal configuration of different performances of the control system and realize the online optimization of compensation parameters:

firstly, combining two obtained optimization objective functions with a Simulink model of a permanent magnet synchronous motor vector control system based on model compensation active disturbance rejection, building a coupling calculation model of a multi-objective optimization algorithm and the Simulink model, and driving the Simulink model to operate by an NSGA-II algorithm; secondly, realizing data interaction of simulation results of the m file and the Simulink model, realizing an online parameter setting process of the model-based compensation active disturbance rejection permanent magnet synchronous motor vector control system by using an NSGA-II algorithm, carrying out repeated iteration solution and taking the average value of a solution set as the optimal value of a compensation parameter; finally, comparing and analyzing simulation results of the Simulink model before optimization, wherein eight calculation processes are adopted for performing parameter online optimization by combining an NSGA-II algorithm based on non-dominated solution sorting and the Simulink model, and are as follows:

the first process is as follows: initializing parameters and randomly generating an initial population P with the size of N₀And setting the Simulink model into motor operation parameters and compensation parameters k_cd、k_cqThe real number coding is carried out in the optimized interval;

and a second process: determining a fitness function, and calculating the fitness value, namely the two selected optimization objective functions are as follows:

in formula (14) f₁(x)，f₂(x) Are all minimized and optimized;

and a third process: driving a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection to operate by an NSGA-II algorithm, performing initial rapid non-dominated sorting and congestion degree calculation, carrying out dominated relation sorting on a first-generation initialized parent-child population and solving a first-level front surface; calculating crowding distances for individuals at the same level and arranging according to the crowding degree at the same level;

and (4) a fourth process: selecting, crossing and varying, merging the father population and the offspring population into 2N individuals, calculating and sequencing fitness values, performing binary tournament with elite strategy to screen out excellent individuals, and selecting from the father population and the offspring population to achieve the optimum under the condition of meeting a certain optimization objective function and not damaging another optimization objective function;

and a fifth process: judging whether the number of generated new populations meets N, otherwise, performing rapid non-dominated sorting and congestion degree calculation again, selecting appropriate individuals to form a new parent population, skipping to the flow four to perform fitness value calculation again and sorting, and finally obtaining new populations of N excellent individuals;

and a sixth process: selecting, crossing and mutating again on the basis of the obtained new population, and further optimizing and screening the population;

a seventh process: judging a termination condition, otherwise, jumping to a third process to generate new parent and child populations again for iterative operation, wherein the elite strategy for screening the populations of the good individuals needs to meet two conditions, namely: comparing the ranking levels of the non-dominated solutions, the smaller the better, the condition two: if the congestion degree is the same level, the congestion degree is compared, and the larger the congestion degree is, the better the expansibility of the solution is;

and (eight) flow: iterating for multiple times until an optimal solution is found, and outputting a result;

the output result is the compensation parameter k in the vector control of the surface-mounted permanent magnet synchronous motor_cd、k_cqAnd (5) optimizing the final result.

Technical Field

The invention relates to a permanent magnet synchronous motor multi-target parameter optimization method based on an entropy weight method, which is particularly suitable for an online setting and optimization method aiming at control parameters in a Simulink simulation model of a control system in the field of engineering.

Background

The permanent magnet synchronous motor has the advantages of simple structure, high torque current ratio, high power density and the like, and is widely applied to the field of power driving. The vector control of the permanent magnet synchronous motor generally comprises a rotating speed loop of an outer loop, a current loop of an inner loop and a PWM control algorithm. The rotating speed loop and the current loop are usually PID control, Sliding Mode Control (SMC), proportional resonant control (PR), Active Disturbance Rejection Control (ADRC), Model Predictive Control (MPC) and the like. And each loop control needs parameter setting and optimizing work. Theoretically, for an actual control system which cannot accurately express a transfer function, parameter setting and optimization work cannot be performed by using a pole allocation method, an amplitude-phase margin method, a mode identification method and the like of a classical control theory. Therefore, in engineering, the parameters are usually adjusted by manual experience depending on a critical proportionality (ZN method), an attenuation curve method, an experience trial and error method, and the parameters of the controller are various and the adjustment amount is large.

In the method for online setting and optimizing parameters, a single control performance is mostly used as an optimization target, and different performance indexes are selected, so that parameters, structures and the like of the controller are different, when the optimal controller is designed, multi-objective optimization of the control performance is always considered, contradictions among the multiple control performance indexes exist all the time, and if the contradictions exist between overshoot and response time, the reduction of the overshoot will cause the increase of the response time. The contradiction between the steady-state error and the adjustment time, and the decrease of the steady-state error will cause the increase of the adjustment time, etc. Or aiming at parameter setting and optimization of multiple control performance indexes, a penalty function or a weighting processing method is generally adopted to subjectively assign weights, and further simplification is realized as a single-target optimization problem. NeedleIn order to solve the problems, the invention provides an entropy weight method-based multi-target NSGA-II parameter setting and optimizing method for a permanent magnet synchronous motor, which is used for compensating a parameter k for a model in a control system_cd、k_cqAnd online setting and optimization are carried out, and the rotation speed of the permanent magnet synchronous motor is mainly realized to have smaller overshoot, faster response speed and good steady-state performance.

Disclosure of Invention

The invention aims to solve the problems that most of motor control parameter setting optimization is single-target optimization and weight is subjectively distributed under multiple control performances in the prior art, and provides a permanent magnet synchronous motor multi-target parameter optimization method based on an entropy weight method.

In order to achieve the above purpose, the technical solution of the invention is as follows:

a permanent magnet synchronous motor multi-target parameter optimization method based on an entropy weight method comprises the following steps:

the first step is as follows: establishing a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection,

the controlled object is a mathematical model of the surface-mounted permanent magnet synchronous motor, and is shown in formula (1):

u in formula (1)_dStator voltage of d-axis, u_qStator voltage of q-axis, i_dStator voltage of d-axis, i_qStator current of q-axis, L_dStator inductance of d-axis, L_qStator inductance of q axis, R stator phase resistance, w_eIs the mechanical angular velocity, psi, of the motor_fIs a mechanical angle flux linkage of the motor, P_nIs the magnetic pole pair number, J is the rotational inertia of the motor, B is the damping coefficient of the motor, T_eIs the electromagnetic torque of the machine, T_LIs the load torque of the motor;

the surface-mounted permanent magnet synchronous motor adopts_d ^*A 0 double closed-loop vector control strategy, wherein a space vector pulse is applied in the vector controlThe wide modulation technology is used for carrying out frequency conversion speed regulation on the motor, the rotating speed ring carries out model compensation on a second-order linear active disturbance rejection controller by using a sliding mode disturbance observer, and the formula (2) is as follows:

in the formula (2), y is the system rotating speed output w_e，z₁Differential tracking signal of y, z₂Is z₁Of the differential tracking signal, z₃For differential tracking signals of system disturbances, e is w_eβ, β₁Gain corrected for zero order of output error, β₂To output a first order correction gain for the error, β₃Is the second-order correction gain of the output error, u is the second-order linear active disturbance rejection controller rotational speed output of the model compensation,is the estimated output of the sliding mode disturbance observer;

the current loop in the vector control of the surface-mounted permanent magnet synchronous motor adopts PI control, and in addition, the estimated output of a sliding mode disturbance observer is applied

Actual output value w of rotating speed_eD-axis current actual value i_dAnd the actual value of q-axis current i_qDesigning a compensation control strategy of a current loop; the compensation control strategy aiming at the current loop adopts the method of combining the feedforward decoupling theory and setting the compensation parameter k_cd、k_cqTo perform feed-forward compensation on the current loop, as shown in equation (3):

k in formula (3)_pdProportional gain, K, of d-axis current loop PI controller_pqProportional gain, K, of a q-axis current loop PI controller_idD-axis current loop PI controller integral gain, K_iqCurrent loop PI controller integral increase for q-axisYi, i_d ^*Current loop command input value, i, for d-axis_q ^*Is the current loop instruction input value of the q axis, s is a Laplace operator,for the d-axis current loop voltage compensation value,

a current loop voltage compensation value for the q axis;

the key point in the feedforward compensation control strategy is to use the estimated value of the sliding mode disturbance observer

(i.e., disturbance estimate) is introduced into the compensation control of the current loop and a compensation parameter k is set_cd、k_cqAdjusting, wherein the specific design expression is shown as a formula (4);

k in the formula (4)_cdCurrent loop compensation parameter, k, for d-axis_cqCurrent loop compensation parameters for the q-axis;

therefore, a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection is built through the design of a rotating speed loop and a current loop and the related space vector pulse width modulation technology;

the second step is that: carrying out multi-target layering on a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection by using a comprehensive evaluation method and establishing a corresponding evaluation index system;

the establishment of the evaluation index system comprises the steps of taking the control system with the aspects of overshoot, response rapidity, stationarity and accuracy as a target decision layer;

respectively confirming index parameters influencing the objective decision layer: by peak time t_pRising time t_rAnd adjusting the time t_sIn order to characterize the indicator of rapidity of response,the method comprises the following steps of (1) taking a rotating speed steady-state error as an accuracy index represented, taking a rotating speed volatility tau and a rotating speed residual error e as stability indexes represented, and taking overshoot as overshoot performance indexes represented; seven performance indexes representing four aspects of the system are used as index layers;

the specific calculation of the index layer is that on the basis of the Simulink model established in the first step, a target rotating speed w is set_r1000r/min, actual output value w of rotating speed when starting under no-load condition_eSeven indices were calculated as follows:

i rapidity index: time of peak t_pRising time t_rAdjusting the time t_sTime of peak t_pTime taken for the first time to reach the maximum output of the rotation speed waveform, rise time t_rAdjusting the time t for the first time to reach the set rotation speed_sAdjusting the time t for the time when the set rotation speed allowable error range is reached for the first time in the steady state process_sThe allowable error range of (2%);

ii accuracy index: the rotating speed steady-state error is the difference value between the rotating speed output value and the rotating speed average value in the steady-state process, and the calculation formula is shown as the formula (5):

＝w_e-w_{e_mean}(5)

iii smoothness index: the rotation speed fluctuation tau is the ratio of the maximum and minimum difference values of the rotation speed in the steady state process, and the calculation formula is shown as the formula (6):

w in the formulae (5) and (6)_eActual output value of the rotational speed at no-load start, w_{e_max}Is the actual output value w of the rotating speed in the steady state process_eMaximum value of, w_{e_min}Is the actual output value w of the rotating speed in the steady state process_eMinimum value of, w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the rotating speed residual e is the difference value between the output value of the rotating speed and the set value of the rotating speed, and the calculation formula is shown as the formula (7):

e＝w_e-w_r(7)

iv, overshooting performance indexes: overshoot, overshoot being the peak time t_pThe calculation formula of the ratio of the maximum deviation value of the lower rotating speed output to the output value in the steady state process is shown as the formula (8):

w in the formulae (7) and (8)_rTarget rotational speed set for the control system (i.e. 1000r/min), w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the third step: regularizing the data matrixes with different performance indexes in the index evaluation system established in the second step, reasonably distributing the weight of each index to multiple indexes with different performances, and preferentially processing the overshoot if the overshoot problem exists in the control system, so that penalty function processing is performed on the overshoot, and entropy weight processing is performed on the multiple indexes with the weight needing to be distributed;

the overshoot is subject to a penalty function, if overshoot exists, the coefficient given by the index is far larger than the coefficients of other indexes, an entropy weight method is adopted to objectively distribute weight values for three indexes (namely, rotational speed volatility tau, rotational speed residual error e and steady-state error) for representing the stability and the accuracy of a control system, and finally the indexes of the rotational speed volatility tau, the rotational speed residual error e, the steady-state error and the overshoot are integrated into an optimized objective function, wherein the calculation formula is shown as formula (9):

g₁(t)＝w₁·τ+w₂·e+w₃·+c· (9)

w in formula (9)₁For objective assignment of the weight, w, of the rotational speed ripple τ₂For objective distribution of weight sum w of rotation speed residual e₃For objective assignment of weight, w, of steady-state error₁、w₂And w₃The sum is equal to 1, c is a penalty factor and is far more than 0, c is 100, g₁(t) is a first optimization objective function;

in the evaluation index system, three indexes in seven indexes are as follows: the rotation speed volatility tau, the rotation speed residual error e and the steady-state error are distributed with weight by adopting an entropy weight method, and the three calculation processes of the entropy weight method are as follows:

the first process is as follows: and (3) data standardization, namely standardizing each index data:

in the formula (10), i is an index data sequence, j is an index number, and x_{i_max}Is the maximum value of j index, x_{i_min}Is the minimum value of the j index;

and a second process: calculating the information entropy of each index:

in the formula (11)Is the proportion of single data in the j target lower index;

and a third process: determining the weight of each index:

in the formula (12), j is 1,2,3, …, and n is the weight of each index;

the peak time t for representing the response rapidity in the evaluation index system_pRising time t_rAdjusting the time t_sThe three indexes are all related to time, so the time minimization principle is adopted for design; and combining a penalty function of the overshoot to emphatically prevent the control system from generating overshoot; therefore, the peak time t_pRising time t_rAdjusting the time t_sAnd the indexes of overshoot are integrated into another optimization objective function, and the calculation formula is shown as the formula (13):

g(t)＝t_p+t_r+t_s+c· (13)

in formula (13), g₂(t) is a second optimization objective function, c is a penalty factor and is far greater than 0, and the value of c is 100;

therefore, the established evaluation index system is quantized to obtain two optimized objective functions: the first is an optimization objective function related to smooth and accurate output of the rotation speed, see equation (9), and the second is an optimization objective function related to time (i.e., response rapidity), see equation (13);

the fourth step: applying the two optimized objective functions obtained in the third step to a compensation parameter k in a Simulink model by using a non-dominated solution sorting-based NSGA-II algorithm_cd、k_cqOptimizing so as to realize the optimal configuration of different performances of the control system and realize the online optimization of compensation parameters:

firstly, combining two obtained optimization objective functions with a Simulink model of a permanent magnet synchronous motor vector control system based on model compensation active disturbance rejection, building a coupling calculation model of a multi-objective optimization algorithm and the Simulink model, and driving the Simulink model to operate by an NSGA-II algorithm; secondly, realizing data interaction of simulation results of the m file and the Simulink model, realizing an online parameter setting process of the model-based compensation active disturbance rejection permanent magnet synchronous motor vector control system by using an NSGA-II algorithm, carrying out repeated iteration solution and taking the average value of a solution set as the optimal value of a compensation parameter; finally, comparing and analyzing simulation results of the Simulink model before optimization, wherein eight calculation processes are adopted for performing parameter online optimization by combining an NSGA-II algorithm based on non-dominated solution sorting and the Simulink model, and are as follows:

the first process is as follows: initializing parameters and randomly generating an initial population P with the size of N₀And setting the Simulink model into motor operation parameters and compensation parameters k_cd、k_cqThe real number coding is carried out in the optimized interval;

and a second process: determining a fitness function, and calculating the fitness value, namely the two selected optimization objective functions are as follows:

in formula (14) f₁(x)，f₂(x) Are all minimized and optimized;

and a third process: driving a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection to operate by an NSGA-II algorithm, performing initial rapid non-dominated sorting and congestion degree calculation, carrying out dominated relation sorting on a first-generation initialized parent-child population and solving a first-level front surface; calculating crowding distances for individuals at the same level and arranging according to the crowding degree at the same level;

and (4) a fourth process: selecting, crossing and varying, merging the father population and the offspring population into 2N individuals, calculating and sequencing fitness values, performing binary tournament with elite strategy to screen out excellent individuals, and selecting from the father population and the offspring population to achieve the optimum under the condition of meeting a certain optimization objective function and not damaging another optimization objective function;

and a fifth process: judging whether the number of generated new populations meets N, otherwise, performing rapid non-dominated sorting and congestion degree calculation again, selecting appropriate individuals to form a new parent population, skipping to the flow four to perform fitness value calculation again and sorting, and finally obtaining new populations of N excellent individuals;

and a sixth process: selecting, crossing and mutating again on the basis of the obtained new population, and further optimizing and screening the population;

a seventh process: judging a termination condition, otherwise, jumping to a third process to generate new parent and child populations again for iterative operation, wherein the elite strategy for screening the populations of the good individuals needs to meet two conditions, namely: comparing the ranking levels of the non-dominated solutions, the smaller the better, the condition two: if the congestion degree is the same level, the congestion degree is compared, and the larger the congestion degree is, the better the expansibility of the solution is;

and (eight) flow: iterating for multiple times until an optimal solution is found, and outputting a result;

the output result is the compensation parameter k in the vector control of the surface-mounted permanent magnet synchronous motor_cd、k_cqAnd (5) optimizing the final result.

Compared with the prior art, the invention has the beneficial effects that:

1. the invention relates to a permanent magnet synchronous motor multi-objective parameter optimization method based on an entropy weight method, which is based on the combination of computer simulation analysis and a numerical simulation technology and aims at carrying out parameter online setting and optimization on an actual control system which is difficult to express a transfer function. And when a comprehensive evaluation method is used for establishing the multi-objective optimization function, the calculation formula adopted according to the optimization target direction is richer than the previous parameter optimization, and only the generalized integral of the output error is used as the optimization calculation formula. And selecting the overshoot, the stability, the rapidity and the accuracy of the slave control system, respectively establishing corresponding index calculation formulas, and further serving as the basis of a multi-objective optimization function. Therefore, the design adopts a relatively objective method to carry out multi-objective parameter optimization, and can further optimize the overall control of the motor.

2. The invention relates to a permanent magnet synchronous motor multi-target parameter optimization method based on an entropy weight method, which is characterized in that MATLAB/Simulink is used for carrying out simulation modeling on a permanent magnet synchronous motor vector control system based on model compensation active disturbance rejection, and a basis is provided for NSGA-II algorithm optimization based on non-dominated solution sorting. Aiming at the problem of subjective distribution of the weights of a plurality of performance indexes, an entropy weight method is adopted to objectively distribute the weights, and subjective factors of parameter setting and optimization are well avoided. The non-dominated sorting genetic algorithm with the elite strategy is introduced into the control parameter setting technology of the permanent magnet synchronous motor, so that the limitation of the conventional single target of parameter optimization is eliminated, and an effective method for multi-target optimization is provided for the on-line setting and optimization of the control parameters of the Simulink model. Therefore, the design evaluation objectively and effectively balances the relation among the control performances, so that the motor can quickly respond and maintain a stable state.

3. The invention relates to a permanent magnet synchronous motor multi-objective parameter optimization method based on an entropy weight method, which adopts a Simulink model and a multi-objective optimization algorithm to combine and iterate to obtain optimal parameters, and provides a new idea for the technical field of control parameter setting of permanent magnet synchronous motors. Firstly, the NSGA-II algorithm based on non-dominated solution sorting and the Simulink model designed by the invention are used for carrying out online optimization on control parameters for multiple times, then parameter adjustment is carried out on the simulated actual working condition by means of the parameter optimal solution, and finally, the control parameters are directly set on an engineering system and run to obtain good control performance. Therefore, the invention brings great convenience to the complicated work of debugging the parameters depending on experience in the past, improves the comprehensive performance of the system, reduces the time and cost of parameter on-line setting, improves the parameter setting efficiency and the parameter accuracy, and has certain social and economic benefits.

4. In the multi-objective parameter optimization method of the permanent magnet synchronous motor based on the entropy weight method, two optimization objectives are selected and only two compensation parameters are set and optimized on line; if the computer operation condition allows, a plurality of targets can be realized to carry out on-line setting and optimization work aiming at all the parameters to be adjusted, thereby realizing the control and debugging work without parameterization. Therefore, the design can provide a preliminary design idea for the control parameter self-adaption of the subsequent motor.

Drawings

FIG. 1 is a flow chart of a permanent magnet synchronous motor multi-objective parameter optimization method based on an entropy weight method.

Fig. 2 is a block diagram of the vector control of the permanent magnet synchronous motor based on model compensation active disturbance rejection.

FIG. 3 is a simulation diagram of the vector control of the PMSM based on model-compensated active-disturbance-rejection in the invention.

FIG. 4 is a block diagram of the tachometer ring model compensation LADRC control of FIG. 2 in accordance with the present invention.

FIG. 5 is a system diagram of evaluation indexes in the comprehensive evaluation method of the present invention.

FIG. 6 is a flow chart of the NSGA-II algorithm in combination with the Simulink model for parameter online optimization.

FIG. 7 is a graph of the results of the NSGA-II algorithm of the present invention.

FIG. 8 is a graph of the rapid performance of the invention before and after optimization.

FIG. 9 is a graph comparing the smooth performance before and after optimization according to the present invention.

Detailed Description

The present invention will be described in further detail with reference to the following description and embodiments in conjunction with the accompanying drawings.

Referring to fig. 1 to 9, a method for optimizing multi-objective parameters of a permanent magnet synchronous motor based on an entropy weight method includes the following steps:

the first step is as follows: establishing a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection, wherein a controlled object is a mathematical model of a surface-mounted permanent magnet synchronous motor, and the mathematical model is as shown in formula (1):

u in formula (1)_dStator voltage of d-axis, u_qStator voltage of q-axis, i_dStator voltage of d-axis, i_qStator current of q-axis, L_dStator inductance of d-axis, L_qStator inductance of q axis, R stator phase resistance, w_eIs the mechanical angular velocity, psi, of the motor_fIs a mechanical angle flux linkage of the motor, P_nIs the magnetic pole pair number, J is the rotational inertia of the motor, B is the damping coefficient of the motor, T_eIs the electromagnetic torque of the machine, T_LIs the load torque of the motor;

the surface-mounted permanent magnet synchronous motor adopts_d ^*The method comprises the following steps that (1) a double closed-loop vector control strategy is set to 0, a space vector pulse width modulation technology is applied in vector control to carry out variable frequency speed regulation on a motor, a sliding mode disturbance observer is utilized to carry out model compensation on a second-order linear active disturbance rejection controller in a rotating speed loop, and the formula (2) is as follows:

in the formula (2), y is the system rotating speed output w_e，z₁Differential tracking signal of y, z₂Is z₁Of the differential tracking signal, z₃For differential tracking signals of system disturbances, e is w_eβ, β₁Gain corrected for zero order of output error, β₂To output a first order correction gain for the error, β₃Is the second-order correction gain of the output error, u is the second-order linear active disturbance rejection controller rotational speed output of the model compensation,is the estimated output of the sliding mode disturbance observer;

the current loop in the vector control of the surface-mounted permanent magnet synchronous motor adopts PI control, and in addition, the estimated output of a sliding mode disturbance observer is appliedActual output value w of rotating speed_eD-axis current actual value i_dAnd the actual value of q-axis current i_qDesigning a compensation control strategy of a current loop; the compensation control strategy aiming at the current loop adopts the method of combining the feedforward decoupling theory and setting the compensation parameter k_cd、k_cqTo perform feed-forward compensation on the current loop, as shown in equation (3):

k in formula (3)_pdProportional gain, K, of d-axis current loop PI controller_pqProportional gain, K, of a q-axis current loop PI controller_idD-axis current loop PI controller integral gain, K_iqCurrent loop PI controller integral gain, i, for q-axis_d ^*Current loop command input value, i, for d-axis_q ^*Is the current loop instruction input value of the q axis, s is a Laplace operator,for the d-axis current loop voltage compensation value,a current loop voltage compensation value for the q axis;

the key point in the feedforward compensation control strategy is to use the estimated value of the sliding mode disturbance observer(i.e., disturbance estimate) is introduced into the compensation control of the current loop and a compensation parameter k is set_cd、k_cqAdjusting, wherein the specific design expression is shown as a formula (4);

k in the formula (4)_cdCurrent loop compensation parameter, k, for d-axis_cqCurrent loop compensation parameters for the q-axis;

therefore, a permanent magnet synchronous motor vector control Simulink model based on model compensation active disturbance rejection is built through the design of a rotating speed loop and a current loop and the related space vector pulse width modulation technology;

the second step is that: carrying out multi-target layering on a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection by using a comprehensive evaluation method and establishing a corresponding evaluation index system;

the establishment of the evaluation index system comprises the steps of taking the control system with the aspects of overshoot, response rapidity, stationarity and accuracy as a target decision layer;

respectively confirming index parameters influencing the objective decision layer: by peak time t_pRising time t_rAnd adjusting the time t_sThe method is characterized by comprising the following steps of (1) taking a response rapidness index as a characteristic, taking a rotating speed steady-state error as an accuracy index as a characteristic, taking a rotating speed volatility tau and a rotating speed residual error e as a stability index as a characteristic, and taking an overshoot as an overshoot performance index as a characteristic; seven performance indexes representing four aspects of the system are used as index layers;

the specific calculation of the index layer is that on the basis of the Simulink model established in the first step, a target rotating speed w is set_r1000r/min, actual output value w of rotating speed when starting under no-load condition_eSeven indices were calculated as follows:

i rapidity index: time of peak t_pRising time t_rAdjusting the time t_sTime of peak t_pIs as followsTime taken for one time to reach maximum output value of rotating speed waveform, rising time t_rAdjusting the time t for the first time to reach the set rotation speed_sAdjusting the time t for the time when the set rotation speed allowable error range is reached for the first time in the steady state process_sThe allowable error range of (2%);

ii accuracy index: the rotating speed steady-state error is the difference value between the rotating speed output value and the rotating speed average value in the steady-state process, and the calculation formula is shown as the formula (5):

＝w_e-w_{e_mean}(5)

iii smoothness index: the rotation speed fluctuation tau is the ratio of the maximum and minimum difference values of the rotation speed in the steady state process, and the calculation formula is shown as the formula (6):

w in the formulae (5) and (6)_eActual output value of the rotational speed at no-load start, w_{e_max}Is the actual output value w of the rotating speed in the steady state process_eMaximum value of, w_{e_min}Is the actual output value w of the rotating speed in the steady state process_eMinimum value of, w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the rotating speed residual e is the difference value between the output value of the rotating speed and the set value of the rotating speed, and the calculation formula is shown as the formula (7):

e＝w_e-w_r(7)

iv, overshooting performance indexes: overshoot, overshoot being the peak time t_pThe calculation formula of the ratio of the maximum deviation value of the lower rotating speed output to the output value in the steady state process is shown as the formula (8):

w in the formulae (7) and (8)_rTarget rotational speed set for the control system (i.e. 1000r/min), w_{e_mean}Is the actual output value w of the rotating speed in the steady state process_eAverage value of (d);

the third step: regularizing the data matrixes with different performance indexes in the index evaluation system established in the second step, reasonably distributing the weight of each index to multiple indexes with different performances, and preferentially processing the overshoot if the overshoot problem exists in the control system, so that penalty function processing is performed on the overshoot, and entropy weight processing is performed on the multiple indexes with the weight needing to be distributed;

the overshoot is subject to a penalty function, if overshoot exists, the coefficient given by the index is far larger than the coefficients of other indexes, an entropy weight method is adopted to objectively distribute weight values for three indexes (namely, rotational speed volatility tau, rotational speed residual error e and steady-state error) for representing the stability and the accuracy of a control system, and finally the indexes of the rotational speed volatility tau, the rotational speed residual error e, the steady-state error and the overshoot are integrated into an optimized objective function, wherein the calculation formula is shown as formula (9):

g₁(t)＝w₁·τ+w₂·e+w₃·+c· (9)

w in formula (9)₁For objective assignment of the weight, w, of the rotational speed ripple τ₂For objective distribution of weight sum w of rotation speed residual e₃For objective assignment of weight, w, of steady-state error₁、w₂And w₃The sum is equal to 1, c is a penalty factor and is far more than 0, c is 100, g₁(t) is a first optimization objective function;

in the evaluation index system, three indexes in seven indexes are as follows: the rotation speed volatility tau, the rotation speed residual error e and the steady-state error are distributed with weight by adopting an entropy weight method, and the three calculation processes of the entropy weight method are as follows:

the first process is as follows: and (3) data standardization, namely standardizing each index data:

in the formula (10), i is an index data sequence, j is an index number, and x_{i_max}Is the maximum value of j index, x_{i_min}Is the minimum value of the j index;

and a second process: calculating the information entropy of each index:

in the formula (11)

Is the proportion of single data in the j target lower index;

and a third process: determining the weight of each index:

in the formula (12), j is 1,2,3, …, and n is the weight of each index;

the peak time t for representing the response rapidity in the evaluation index system_pRising time t_rAdjusting the time t_sThe three indexes are all related to time, so the time minimization principle is adopted for design; and combining a penalty function of the overshoot to emphatically prevent the control system from generating overshoot; therefore, the peak time t_pRising time t_rAdjusting the time t_sAnd the indexes of overshoot are integrated into another optimization objective function, and the calculation formula is shown as the formula (13):

g(t)＝t_p+t_r+t_s+c· (13)

in formula (13), g₂(t) is a second optimization objective function, c is a penalty factor and is far greater than 0, and the value of c is 100;

therefore, the established evaluation index system is quantized to obtain two optimized objective functions: the first is an optimization objective function related to smooth and accurate output of the rotation speed, see equation (9), and the second is an optimization objective function related to time (i.e., response rapidity), see equation (13);

the fourth step: applying the two optimized objective functions obtained in the third step to a compensation parameter k in a Simulink model by using a non-dominated solution sorting-based NSGA-II algorithm_cd、k_cqOptimizing so as to realize the optimal configuration of different performances of the control system and realize the online optimization of compensation parameters:

firstly, combining two obtained optimization objective functions with a Simulink model of a permanent magnet synchronous motor vector control system based on model compensation active disturbance rejection, building a coupling calculation model of a multi-objective optimization algorithm and the Simulink model, and driving the Simulink model to operate by an NSGA-II algorithm; secondly, realizing data interaction of simulation results of the m file and the Simulink model, realizing an online parameter setting process of the model-based compensation active disturbance rejection permanent magnet synchronous motor vector control system by using an NSGA-II algorithm, carrying out repeated iteration solution and taking the average value of a solution set as the optimal value of a compensation parameter; finally, comparing and analyzing simulation results of the Simulink model before optimization, wherein eight calculation processes are adopted for performing parameter online optimization by combining an NSGA-II algorithm based on non-dominated solution sorting and the Simulink model, and are as follows:

the first process is as follows: initializing parameters and randomly generating an initial population P with the size of N₀And setting the Simulink model into motor operation parameters and compensation parameters k_cd、k_cqThe real number coding is carried out in the optimized interval;

and a second process: determining a fitness function, and calculating the fitness value, namely the two selected optimization objective functions are as follows:

in formula (14) f₁(x)，f₂(x) Are all minimized and optimized;

and a third process: driving a permanent magnet synchronous motor vector control system Simulink model based on model compensation active disturbance rejection to operate by an NSGA-II algorithm, performing initial rapid non-dominated sorting and congestion degree calculation, carrying out dominated relation sorting on a first-generation initialized parent-child population and solving a first-level front surface; calculating crowding distances for individuals at the same level and arranging according to the crowding degree at the same level;

and (4) a fourth process: selecting, crossing and varying, merging the father population and the offspring population into 2N individuals, calculating and sequencing fitness values, performing binary tournament with elite strategy to screen out excellent individuals, and selecting from the father population and the offspring population to achieve the optimum under the condition of meeting a certain optimization objective function and not damaging another optimization objective function;

and a fifth process: judging whether the number of generated new populations meets N, otherwise, performing rapid non-dominated sorting and congestion degree calculation again, selecting appropriate individuals to form a new parent population, skipping to the flow four to perform fitness value calculation again and sorting, and finally obtaining new populations of N excellent individuals;

and a sixth process: selecting, crossing and mutating again on the basis of the obtained new population, and further optimizing and screening the population;

a seventh process: judging a termination condition, otherwise, jumping to a third process to generate new parent and child populations again for iterative operation, wherein the elite strategy for screening the populations of the good individuals needs to meet two conditions, namely: comparing the ranking levels of the non-dominated solutions, the smaller the better, the condition two: if the congestion degree is the same level, the congestion degree is compared, and the larger the congestion degree is, the better the expansibility of the solution is;

and (eight) flow: iterating for multiple times until an optimal solution is found, and outputting a result;

the output result is the compensation parameter k in the vector control of the surface-mounted permanent magnet synchronous motor_cd、k_cqAnd (5) optimizing the final result.

The principle of the invention is illustrated as follows:

the learner Yongting ding already teaches a design flow of a sliding mode disturbance observer in ISA Adaptive sliding mode current control with sliding mode disturbance for PMSM drives, which corresponds to a model building part of a vector control system Simulink of the permanent magnet synchronous motor based on model compensation active disturbance in the first step, wherein the sliding mode disturbance observer design flow is shown in fig. 2, and therefore, the patent of the present invention is not repeated herein.

Simulink: the system is a visual simulation tool in MATLAB, is a block diagram design environment based on MATLAB, is a software package for realizing dynamic system modeling, simulation and analysis, and is widely applied to modeling and simulation of linear systems, nonlinear systems, digital control and digital signal processing. Simulink provides an integrated environment for dynamic system modeling, simulation and comprehensive analysis. In this environment, a complex system can be constructed without a large number of writing programs, but only by simple and intuitive mouse operation.

Space vector pulse width modulation technique (SVPWM): the main idea is that the ideal flux linkage circle of the stator of the three-phase symmetrical motor is used as a reference standard when the three-phase symmetrical sine-wave voltage is used for supplying power, and different switching modes of a three-phase inverter are properly switched, so that PWM waves are formed, and the accurate flux linkage circle is tracked by the formed actual flux linkage vector. The traditional SPWM method starts from the power supply to generate a sine wave power supply with adjustable frequency and voltage, and the SVPWM method considers an inverter system and an asynchronous motor as a whole, so that the model is simple and the real-time control of a microprocessor is facilitated.

Linear Active Disturbance Rejection Controller (LADRC): the Linear active disturbance rejection controller is linearly simplified by an active disturbance rejection controller proposed by hangul, and mainly linearizes an Extended State Observer (ESO) and a feedback control rate in the active disturbance rejection controller, so that the Linear active disturbance rejection controller comprises a second-order Tracking Differentiator (TD), a Linear Extended State Observer (LESO) and a Linear feedback control combination (PD). The linear active disturbance rejection controller solves the problem of complicated parameter adjustment of the active disturbance rejection controller, and simplifies the control model parameters, thereby being more widely applied in the engineering field.

NSGA-II algorithm: NSGA-II is one of the most popular multi-target genetic algorithms, reduces the complexity of the non-inferior ranking genetic algorithm, has the advantages of high running speed and good convergence of solution sets, and becomes the basis of the performance of other multi-target optimization algorithms.