WO2020155509A1 - 考虑铁损的永磁同步电机随机命令滤波神经网络控制方法 - Google Patents
考虑铁损的永磁同步电机随机命令滤波神经网络控制方法 Download PDFInfo
- Publication number
- WO2020155509A1 WO2020155509A1 PCT/CN2019/090499 CN2019090499W WO2020155509A1 WO 2020155509 A1 WO2020155509 A1 WO 2020155509A1 CN 2019090499 W CN2019090499 W CN 2019090499W WO 2020155509 A1 WO2020155509 A1 WO 2020155509A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- function
- formula
- error
- inequality
- above formula
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P6/00—Arrangements for controlling synchronous motors or other dynamo-electric motors using electronic commutation dependent on the rotor position; Electronic commutators therefor
- H02P6/34—Modelling or simulation for control purposes
Definitions
- the invention belongs to the technical field of permanent magnet synchronous motor position tracking control, and in particular relates to a permanent magnet synchronous motor random command filtering neural network control method considering iron loss.
- PMSM permanent magnet synchronous motors
- the PMSM system is highly non-linear, strongly coupled and multivariable, and the motor system will be disturbed by some uncertain factors in practical applications, such as parameter uncertainties and load disturbances.
- related scientific and technological workers have proposed some advanced nonlinear control methods and achieved good results, such as advanced control technologies such as backstepping control, sliding mode control, and robust control.
- the motor system can achieve better performance.
- the above-mentioned control method rarely considers the core loss and the influence of random disturbance during the operation of the motor.
- the motor system In the actual application of the motor, the motor system often has a certain core loss due to high-speed operation, no-load operation or light-load operation, which will reduce the stability of the motor system to a certain extent, such as in the emerging hybrid vehicles In the vehicle drive system, the core loss during operation cannot be avoided, so it is extremely important to solve the problem of core loss.
- the motor system cannot avoid random disturbances in practical applications, such as random switching of external loads, random noise and vibration, etc.
- some parameters of the PMSM system will be affected to a certain extent due to random disturbances such as damping torque.
- the adaptive backstepping method has been successfully applied to the PMSM system, but its use has limitations.
- the object it uses must be a linear system, and secondly in the virtual control function
- the problem of "calculation explosion" will appear during the calculation process.
- approximate theories such as fuzzy logic system (FLS) or neural network (NN) have been proposed, and some of the shortcomings in the traditional adaptive backstepping method have been successfully solved, but these methods have not solved the "computational explosion” problem.
- FLS fuzzy logic system
- N neural network
- DSC dynamic surface control
- the DCS technology cannot eliminate the errors generated in actual use, which will affect the PMSM control precision.
- the purpose of the present invention is to propose a permanent magnet synchronous motor random command filtering neural network control method considering iron loss to solve the technical problem of permanent magnet synchronous motor position tracking control under the condition of uncertain parameters and load torque disturbance .
- the random command filtering neural network control method of permanent magnet synchronous motor considering iron loss includes the following steps:
- u d and u q represent stator voltage
- ⁇ and ⁇ represent rotor angular position and rotor angular velocity, respectively
- L d , L q represent stator inductance
- L ld , L lq represent stator leakage inductance
- L md , L mq represent magnetizing inductance
- R 1 represents stator resistance
- R c represents core loss resistance
- T L represents the load torque
- ⁇ PM represents the magnetic flux generated by the permanent magnet of the rotor
- ⁇ 2 , ⁇ 3 , ⁇ 4 , ⁇ 5 , ⁇ 6 all represent unknown smooth nonlinear functions
- L represents the infinite differential operator
- Means Correction term f and h represent the local Lipschitz function of x
- Tr represents the sum of diagonal elements
- u is the symmetric saturated nonlinear input of the permanent magnet synchronous motor system; u can be described as:
- u max > 0 and u min ⁇ 0 are unknown input saturation constants;
- v is a saturated nonlinear input signal;
- D is a constant
- R n represents the set of all n-dimensional real number sequence vectors, and R + represents a positive number;
- ) represents the lower limit of the function V(x), Indicates the upper limit of the function V(x);
- the command filter is defined as follows:
- ⁇ 1 and ⁇ 2 represent real numbers, ⁇ n > 0, ⁇ ⁇ (0, 1];
- q is the input dimension of the neural network
- R q represents the set of all q-dimensional real number sequence vectors
- N * represents the weight vector
- P(Z) represents the basis function vector
- the radial basis function neural network can approximate any continuous function:
- ⁇ (Z) is the tracking error, and satisfies
- N is the unknown ideal weight vector defined in the analysis process, and satisfies
- x 1d is the tracking signal
- x 1,c , x 2,c , x 3,c , x 5,c are the input signals ⁇ 1 , ⁇ 2 , ⁇ 3 , and ⁇ 5 , respectively.
- N j represents the unknown ideal weight vector, among them, Represents the estimated value of ⁇ j;
- P i (Z) P i represents a basis function vector is, ⁇ i (Z) of the tracking error, and satisfies
- Each step of the control method design will select a Lyapunov function to construct a virtual control function or real control law.
- the design of the control method specifically includes the following steps:
- k 1 represents a positive design parameter
- d represents the upper limit of
- l 3 is the selected normal number, which can be obtained:
- k 3 represents a positive design parameter
- the compensation signal is: Select the actual control input as:
- k 4 represents a positive design parameter, and d q is a positive number
- D q is a constant
- l 5 is the selected normal number, which can be obtained:
- k 5 represents a positive design parameter, which can be obtained:
- d d represents a constant
- D d represents a constant
- d 2 is a constant
- Lyapunov function is selected as follows:
- p j and ⁇ j both represent design constants
- a 0 min ⁇ 4e 1 ,4e 2 ,4e 3 ,4e 4 ,4e 5 ,4e 6 ,p 2 ,p 3 ,p 4 ,p 5 ,p 6 ⁇ ;
- the formula (36) can be expressed as: LV ⁇ -a 0 V+b 0 , t ⁇ 0;
- Equation (36) can be expressed as:
- V ⁇ is uniformly bounded, and the compensation signal ⁇ i is also bounded.
- the present invention takes into account the core loss and random interference during the operation of the permanent magnet synchronous motor system, so that the designed control method is more in line with actual engineering needs.
- the present invention is aimed at the highly nonlinear system in the permanent magnet synchronous motor drive system.
- the present invention uses neural network approximation method to deal with, and effectively solves the input saturation problem by using the piecewise smooth function approximation method.
- the present invention effectively solves the "calculation explosion" problem in the calculation process of the random nonlinear system of permanent magnet synchronous motor considering iron loss by using command filtering technology, and compensates the generated filtering error through the command filtering compensator.
- the controller proposed by the present invention is easier to implement than a motor system using a dynamic surface.
- the results prove that the command filter controller can better and faster track the position signal and has stronger robustness .
- Figure 1 is a schematic diagram of a composite controlled object composed of a permanent magnet synchronous motor random command filtering neural network controller, coordinate transformation, and SVPWM inverter considering iron loss in the present invention
- Fig. 2 is a simulation diagram of tracking the rotor angle and the rotor angle setting value after adopting the control method of the present invention
- FIG. 3 is a simulation diagram of the tracking error of the rotor angle and the rotor angle setting value after the control method of the present invention is adopted;
- FIG. 4 is a simulation diagram of the d-axis stator voltage of a synchronous motor after adopting the control method of the present invention
- Fig. 5 is a simulation diagram of the q-axis stator voltage of a synchronous motor after adopting the control method of the present invention.
- the basic idea of the present invention is: use the neural network system to approximate the unknown random nonlinear function in the permanent magnet synchronous motor drive system, and at the same time, based on the Lyapunov function, use the backstepping method to construct the intermediate virtual control signal, and gradually recursively obtain the control law, thereby ensuring The voltage and current are stabilized in a bounded area to reduce errors and improve control accuracy.
- the permanent magnet synchronous motor random command filtering neural network control method considering iron loss mainly includes permanent magnet synchronous motor random command filtering neural network controller 1, coordinate transformation unit 2, and SVPWM inverter 3, rotation speed detection unit 4 and current detection unit 5.
- the speed detection unit 4 and the current detection unit 5 are mainly used to detect the current value and speed-related variables of the permanent magnet synchronous motor.
- the actual measured current and speed variables are used as input, and the permanent magnet synchronous motor random command filtering is taken into account the iron loss.
- the neural network controller 1 performs voltage control and is finally converted into a three-phase electric control permanent magnet synchronous motor's speed. In order to design a more effective controller, it is very important to establish a dynamic model of a permanent magnet synchronous motor.
- the random command filtering neural network control method of permanent magnet synchronous motor considering iron loss includes the following steps:
- u d , u q represent stator voltage; ⁇ and ⁇ represent rotor angular position and rotor angular velocity respectively; i d , i q represent stator current; i od , i oq represent excitation current; J represents moment of inertia; n p represents pole Logarithm; L d , L q represent stator inductance; L ld , L lq represent stator leakage inductance; L md , L mq represent excitation inductance; R 1 represents stator resistance, R c represents core loss resistance; T L represents load rotation Moment; ⁇ PM represents the magnetic flux generated by the permanent magnet of the rotor.
- new variables as follows:
- ⁇ 2 , ⁇ 3 , ⁇ 4 , ⁇ 5 , and ⁇ 6 all represent unknown smooth nonlinear functions.
- L represents the infinite differential operator
- Means The correction term, f and h represent the local Lipschitz function of x; Tr represents the sum of diagonal elements.
- u is the symmetric saturated nonlinear input of the permanent magnet synchronous motor system; u can be described as:
- u max >0 and u min ⁇ 0 are unknown input saturation constants; v is a saturated nonlinear input signal.
- D is a constant.
- R n represents a set of all n-dimensional real number sequence vectors, and R + represents a positive number.
- ) represents the lower limit of the function V(x), Indicates the upper limit of the function V(x).
- V(x) represents the expectation of the function V(x)
- the command filter is defined as follows:
- ⁇ 1 and ⁇ 2 represent real numbers, ⁇ n > 0, and ⁇ ⁇ (0, 1].
- q is the input dimension of the neural network
- R q represents the set of all q-dimensional real number sequence vectors
- N * represents the weight vector
- P(Z) represents the basis function vector
- the radial basis function neural network can approximate any continuous function:
- ⁇ (Z) is the tracking error, and satisfies
- N is the unknown ideal weight vector defined in the analysis process, and satisfies
- x 1d is the tracking signal
- x 1,c , x 2,c , x 3,c , x 5,c are the input signals ⁇ 1 , ⁇ 2 , ⁇ 3 , and ⁇ 5 , respectively.
- N j represents the unknown ideal weight vector, among them, It represents the estimated value of ⁇ j;
- P i (Z) P i represents a basis function vector is, ⁇ i (Z) of the tracking error, and satisfies
- Each step of the control method design will select a Lyapunov function to construct a virtual control function or real control law.
- the design of the control method specifically includes the following steps:
- k 1 represents a positive design parameter.
- d represents the upper limit of
- l 3 is the selected normal number, which can be obtained:
- k 3 represents a positive design parameter
- the compensation signal is: Select the actual control input as:
- k 4 represents a positive design parameter
- d q is a positive number
- D q is a constant.
- l 5 is the selected normal number, which can be obtained:
- k 5 represents a positive design parameter, which can be obtained:
- d d represents a constant.
- D d represents a constant
- d 2 is a constant
- Lyapunov function is selected as follows:
- p j and ⁇ j both represent design constants.
- a 0 min ⁇ 4e 1 ,4e 2 ,4e 3 ,4e 4 ,4e 5 ,4e 6 ,p 2 ,p 3 ,p 4 ,p 5 ,p 6 ⁇ ;
- the formula (36) can be expressed as: LV ⁇ -a 0 V+b 0 , t ⁇ 0.
- Equation (36) can be expressed as:
- V ⁇ is uniformly bounded, and the compensation signal ⁇ i is also bounded.
- the established permanent magnet synchronous motor random command filtering neural network controller with iron loss is simulated to verify the feasibility of the proposed permanent magnet synchronous motor random command filtering neural network method with iron loss:
- the motor and load parameters are:
- the load torque is The control parameters of the filter technology are:
- the simulation is performed under the premise that the system parameters and nonlinear functions are unknown.
- the simulation results of the adaptive neural network control method considering iron loss are shown in the attached figure.
- the tracking signal and the desired signal are shown in Figure 2, and the position tracking error is shown in Figure 3.
- Figure 2 and Figure 3 It can be seen from Figure 2 and Figure 3 that the output of the system can track the desired signal well; the d-axis stator voltage and the q-axis stator voltage are shown in Figure 4 and Figure 5. From Figure 4 and Figure 5, it can be seen that the controller input Both u d and u q are stable in a bounded area.
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Control Of Ac Motors In General (AREA)
- Feedback Control In General (AREA)
Abstract
一种考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,属于永磁同步电机位置跟踪控制技术领域。该方法针对永磁同步电机的控制精度需求以及驱动系统中的随机扰动和非线性问题,在传统的反步法中引入了命令滤波技术来解决计算过程中的"计算爆炸"问题,同时利用神经网络逼近系统中的非线性函数,构造了神经网络自适应位置跟踪控制器。本发明可以保证系统的跟踪误差能够收敛到原点的一个足够小的邻域内,具有更高的工作效率,更强的抗干扰能力以及更好的控制效果。
Description
本发明属于永磁同步电机位置跟踪控制技术领域,尤其涉及一种考虑铁损的永磁同步电机随机命令滤波神经网络控制方法。
近年来,永磁同步电机(PMSM)凭借其结构简单、效率高、使用寿命长和实际运用性强的特点,在农业、工业等领域有着极为广泛的运用。然而,PMSM的系统是高度非线性,强耦合和多变量的,并且在实际运用中电机系统会被一些不确定的因素所干扰,例如参数不确定和负载扰动。为解决这些问题,相关科技工作者提出了一些先进的非线性控制方法并取得了较好的成效,如反步控制、滑模控制、鲁棒控制等先进的控制技术。
通过这些技术可以使电机系统获得更好的性能。然而,上述控制方法很少考虑电机运行过程中的铁芯损耗和随机扰动的影响。在电机的实际运用中,电机系统常常会因为高速运行,空载运行或轻载运行产生一定的铁芯损耗,这将在一定程度上降低电机系统运行的稳定性,例如在新兴的混合动力汽车中,车辆的驱动系统在运行中的铁芯损耗是不能避免的,因此解决铁芯损耗的问题极为重要。另一方面,电机系统在实际运用中也无法避免随机扰动问题,例如外部负载随机切换,随机噪声和振动等,同时,PMSM系统的某些参数会因为阻尼转矩等随机干扰而发生一定程度的变化,这会影响到电机系统运行过程中的控制效果。此外,电机的输入饱和问题也会降低系统运行的稳定性。因此,研究电机运行过程中的铁芯损耗、随机扰动和输入饱和问题对于提高PMSM系统的性能是非常有必要的。
在另一个前端研究领域,作为先进技术之一的自适应反步法已成功运用到了PMSM系统中,但是它的使用却存在局限性,首先它使用的对象必须是线性系统,其次在虚拟控制函数计算过程中会出现“计算爆炸”的问题。针对上述缺点,已经提出了模糊逻辑系统(FLS)或神经网络(NN)等近似理论,并成功解决了传统自适应反步法中存在的部分缺陷,但这些方法并没有解决“计算爆炸”的问题。通过近几年的研究,动态面控制(DSC)技术作为新型技术之一被用来解决计算过程中出现的复杂计算,然而DCS技术在实际使用中不能消除产生的误差,这将影响对PMSM的控制精度。为了改善控制效果,引入了命令滤波误差补偿技术。此外,神经网络系统在处理未知非线性函数方面的能力引起了国内外控制界的广泛关注,并用于具有高度非线性和不确定性的复杂控制系统设计中。
发明内容
本发明的目的在于提出一种考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,以解决在参数不确定和有负载转矩扰动的情况下永磁同步电机的位置跟踪控制的技术问题。
本发明为了实现上述目的,采用如下技术方案:
考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,包括如下步骤:
a建立考虑铁损的永磁同步电机的d-q坐标轴动态数学模型,如公式(1)所示:
其中,u
d,u
q表示定子电压;Θ和ω分别表示转子角位置和转子角速度;
i
d,i
q表示定子电流;i
od,i
oq表示励磁电流;J表示转动惯量;n
p表示极对数;
L
d,L
q表示定子电感;L
ld,L
lq表示定子漏感;L
md,L
mq表示励磁电感;
R
1表示定子电阻,R
c表示铁芯损耗电阻;
T
L表示负载转矩;λ
PM表示转子永磁体产生的磁通量;
为了简化计算过程,定义新的变量如下:
则考虑铁损的永磁同步电机的动态数学模型可表示为:
其中,ψ
2,ψ
3,ψ
4,ψ
5,ψ
6均表示未知的光滑非线性函数;
b基于Lyapunov函数,对于任意给定的函数V=V(x)∈C
2,C
2表示复数集;
定义差分运算,由伊藤微分法则可知:
对于公式(1),u为永磁同步电机系统的对称饱和非线性输入;u可描述为:
其中,u
max>0和u
min<0是未知的输入饱和常数;v是饱和非线性的输入信号;
通过使用分段的平滑函数来近似饱和函数,定义如下函数:
将公式(3)中的sat(v)写成:u=sat(v)=g(v)+d(v);
其中,d(v)=sat(v)-g(v)是有界函数,其边界可表示为:
|d(v)|=|sat(v)-g(v)|≤max{u
max(1-tanh(1)),u
min(tanh(1)-1)}=D (4)
其中,D为常数;
存在常数b
m和b
M,使0<b
m≤g
ia≤b
M<∞成立;
其中,g
ia为x
ia+1的系数,ia=1,…,n;
假设存在一个C
2的函数V(x):R
n→R
+,取两个常数e
0>0和g
0>0,则存在:
其中,R
n表示全体n维实数列向量构成的集合,R
+表示正数;
其中,E[V(x)]表示函数V(x)的期望,V(x
0)表示当x=x
0时V(x)的初始值;
命令滤波器定义如下:
其中,φ
1和φ
2表示实数,ω
n>0,ζ∈(0,1];
其中,ε>0,p>1,q>1,并且(p-1)(q-1)=1;
R
q表示全体q维实数列向量构成的集合,N
*表示权重向量,P(Z)表示基函数向量;
其中,δ(Z)为跟踪误差,且满足|δ(Z)|≤ε;
当N的取值为N
*时,|δ(Z)|达到最小;
下面给出基于反步法的随机非线性神经网络控制器的设计:
定义系统误差变量:
上式中,x
1d为跟踪信号,x
1,c,x
2,c,x
3,c,x
5,c分别为输入信号α
1,α
2,α
3,α
5时滤波器的输出信号,ξ
i为误差补偿信号,i=1,2,3,4,5,6;
定义未知常数θ
j=||N
j||
2,j=2,3,4,5,6,N
j表示未知理想权向量,
其中,
表示θ
j的估计值;P
i(Z)=P
i表示基函数向量,δ
i(Z)为跟踪误差,且满足|δ
i(Z)|≤ε
i;
控制方法设计的每一步都会选取一个Lyapunov函数来构建一个虚拟控制函数或者真实控制律,控制方法的设计具体包括以下步骤:
b1定义误差信号z
1=x
1-x
1d,补偿误差v
1=z
1-ξ
1;
其中,k
1表示正设计参数;
b2定义误差信号z
2=x
2-x
1,c,补偿误差v
2=z
2-ξ
2;
根据杨氏不等式可得:
其中,d表示|T
L|的上限值;d>0;
b3定义误差信号z
3=x
3-x
2,c,补偿误差v
3=z
3-ξ
3;
由杨氏不等式可得:
其中,l
3为选取的正常数,由此可得:
定义x
4=v
4+ξ
4+x
3,c;
b4定义误差信号z
4=x
4-x
3,c,补偿误差v
4=z
4-ξ
4;
其中,l
4为选取的正常数,将该公式代入公式(21)可得:
式中,k
4表示正设计参数,d
q为正数;
由杨氏不等式得到:
其中,D
q为常数;
b5定义误差信号z
5=x
5,补偿误差v
5=z
5-ξ
5;
其中,l
5为选取的正常数,可得:
其中,k
5表示正设计参数,可得到:
b6定义误差信号z
6=x
6-x
5,c,补偿误差v
6=z
6-ξ
6;
由杨氏不等式得:
其中,d
d表示常数;
根据上述公式可知:
其中,D
d表示常数,d
2为常数;
通过上述公式可得:
c由上述分析,选择Lyapunov函数如下:
其中,p
j和λ
j均表示设计常数;
取a
0,b
0分别为:
a
0=min{4e
1,4e
2,4e
3,4e
4,4e
5,4e
6,p
2,p
3,p
4,p
5,p
6};
因此,公式(36)可表示为:LV≤-a
0V+b
0,t≥0;
式(36)可表示为:
上式可改写为:
E[|v
i|
4]≤4E[V(t)],i=1,2,3,4,5,6 (38)
构造Lyapunov函数V
ξ如下:
定义滤波误差为(x
t,c-α
t)且满足|x
t,c-α
t|≤σ
t,t=1,2,3,5,由式(39)可得:
本发明具有如下优点:
(1)本发明考虑了永磁同步电机系统运行过程中铁芯损耗和随机干扰的问题,使设计的控制方法更符合实际工程的需要。
(2)本发明针对永磁同步电机驱动系统中的高度非线性系统,本发明采用神经网络逼近的方法来处理,并通过使用分段光滑函数逼近的方法有效地解决了输入饱和问题。
(3)本发明通过使用命令滤波技术,有效地解决考虑铁损的永磁同步电机随机非线性系统计算过程中的“计算爆炸”问题,并通过命令滤波补偿器对产生的滤波误差进行补偿。
(4)本发明所提出的控制器与使用动态面的电机系统相比更容易实现,结果证明,该命令滤波控制器能够更好更快的实现位置信号的跟踪且具有更强的鲁棒性。
图1本发明中考虑铁损的永磁同步电机随机命令滤波神经网络控制器、坐标变换、SVPWM逆变器组成的复合被控对象的示意图;
图2是采用本发明控制方法后转子角度和转子角度设定值跟踪仿真图;
图3是采用本发明控制方法后转子角度和转子角度设定值跟踪误差仿真图;
图4是采用本发明控制方法后同步电动机d轴定子电压仿真图;
图5是采用本发明控制方法后同步电动机q轴定子电压仿真图。
本发明的基本思想为:利用神经网络系统逼近永磁同步电机驱动系统中未知的随机非线性函数,同时基于Lyapunov函数,运用反步法构造中间虚拟控制信号,逐步递推得到控制律,从而保证电压电流稳定在一个有界区域内,减小误差,提高控制精度。
下面结合附图以及具体实施方式对本发明作进一步详细说明:
如图1所示,考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,其采用的部件主要包括考虑铁损的永磁同步电机随机命令滤波神经网络控制器1、坐标变换单元2、SVPWM逆变器3和转速检测单元4与电流检测单元5。
其中,转速检测单元4和电流检测单元5主要用于检测永磁同步电机的电流值和转速相关变量,通过实际测量的电流和转速变量作为输入,通过考虑铁损的永磁同步电机随机命令滤波神经网络控制器1进行电压控制,最终转换为三相电控制永磁同步电机的转速。为了设计一个更加有效的控制器,建立永磁同步电机动态模型是十分重要的。
考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,包括如下步骤:
a建立考虑铁损的永磁同步电机的d-q坐标轴动态数学模型,如公式(1)所示:
其中,u
d,u
q表示定子电压;Θ和ω分别表示转子角位置和转子角速度;i
d,i
q表示定子电流;i
od,i
oq表示励磁电流;J表示转动惯量;n
p表示极对数;L
d,L
q表示定子电感;L
ld,L
lq表示定子漏感;L
md,L
mq表示励磁电感;R
1表示定子电阻,R
c表示铁芯损耗电阻;T
L表示负载转矩;λ
PM表示转子永磁体产生的磁通量。为了简化计算过程,定义新的变量如下:
则考虑铁损的永磁同步电机的动态数学模型可表示为:
其中,ψ
2,ψ
3,ψ
4,ψ
5,ψ
6均表示未知的光滑非线性函数。
b基于Lyapunov函数,对于任意给定的函数V=V(x)∈C
2,C
2表示复数集;
定义差分运算,由伊藤微分法则可知:
对于公式(1),u为永磁同步电机系统的对称饱和非线性输入;u可描述为:
其中,u
max>0和u
min<0是未知的输入饱和常数;v是饱和非线性的输入信号。
通过使用分段的平滑函数来近似饱和函数,定义如下函数:
将公式(3)中的sat(v)写成:u=sat(v)=g(v)+d(v)。
其中,d(v)=sat(v)-g(v)是有界函数,其边界可表示为:
|d(v)|=|sat(v)-g(v)|≤max{u
max(1-tanh(1)),u
min(tanh(1)-1)}=D (4)
其中,D为常数。
假设存在一个C
2的函数V(x):R
n→R
+,取两个常数e
0>0和g
0>0,则存在:
其中,E[V(x)]表示函数V(x)的期望,V(x
0)表示当x=x
0时V(x)的初始值。
命令滤波器定义如下:
其中,φ
1和φ
2表示实数,ω
n>0,ζ∈(0,1]。
其中,ε>0,p>1,q>1,并且(p-1)(q-1)=1。
其中,δ(Z)为跟踪误差,且满足|δ(Z)|≤ε。
当N的取值为N
*时,|δ(Z)|达到最小。
下面给出基于反步法的随机非线性神经网络控制器的设计:
定义系统误差变量:
上式中,x
1d为跟踪信号,x
1,c,x
2,c,x
3,c,x
5,c分别为输入信号α
1,α
2,α
3,α
5时滤波器的输出信号,ξ
i为误差补偿信号,i=1,2,3,4,5,6。
定义未知常数θ
j=||N
j||
2,j=2,3,4,5,6,N
j表示未知理想权向量,
其中,
表示θ
j的估计值;P
i(Z)=P
i表示基函数向量,δ
i(Z)为跟踪误差,且满足|δ
i(Z)|≤ε
i。
控制方法设计的每一步都会选取一个Lyapunov函数来构建一个虚拟控制函数或者真实控制律,控制方法的设计具体包括以下步骤:
其中,k
1表示正设计参数。
b2定义误差信号z
2=x
2-x
1,c,补偿误差v
2=z
2-ξ
2。
根据杨氏不等式可得:
其中,d表示|T
L|的上限值;d>0。
b3定义误差信号z
3=x
3-x
2,c,补偿误差v
3=z
3-ξ
3。
由杨氏不等式可得:
其中,l
3为选取的正常数,由此可得:
定义x
4=v
4+ξ
4+x
3,c。
b4定义误差信号z
4=x
4-x
3,c,补偿误差v
4=z
4-ξ
4。
其中,l
4为选取的正常数,将该公式代入公式(21)可得:
式中,k
4表示正设计参数,d
q为正数。
由杨氏不等式得到:
其中,D
q为常数。
b5定义误差信号z
5=x
5,补偿误差v
5=z
5-ξ
5。
其中,l
5为选取的正常数,可得:
其中,k
5表示正设计参数,可得到:
b6定义误差信号z
6=x
6-x
5,c,补偿误差v
6=z
6-ξ
6。
由杨氏不等式得:
其中,d
d表示常数。
根据上述公式可知:
其中,D
d表示常数,d
2为常数。
通过上述公式可得:
c由上述分析,选择Lyapunov函数如下:
其中,p
j和λ
j均表示设计常数。
取a
0,b
0分别为:
a
0=min{4e
1,4e
2,4e
3,4e
4,4e
5,4e
6,p
2,p
3,p
4,p
5,p
6};
因此,公式(36)可表示为:LV≤-a
0V+b
0,t≥0。
式(36)可表示为:
上式可改写为:
E[|v
i|
4]≤4E[V(t)],i=1,2,3,4,5,6 (38)
构造Lyapunov函数V
ξ如下:
定义滤波误差为(x
t,c-α
t)且满足|x
t,c-α
t|≤σ
t,t=1,2,3,5,由式(39)可得:
在虚拟环境下对所建立的考虑铁损的永磁同步电机随机命令滤波神经网络控制器进行仿真,验证所提出的考虑铁损的永磁同步电机随机命令滤波神经网络方法的可行性:
电机及负载参数为:
J=0.002kg·m
2,λ
PM=0.0844Wb,R
1=2.21Ω,R
c=200Ω,L
d=0.00977H,L
q=0.00977H,
L
md=0.007H,L
mq=0.008H,L
lq=0.00177H,L
ld=0.00177H,n
p=3。
k
1=k
2=10,k
3=k
4=k
5=k
6=20;l
2=l
3=l
4=l
5=l
6=5;
p
2=p
3=p
4=p
5=p
6=0.5;λ
2=λ
3=λ
4=2,λ
5=0.2,λ
6=2。
命令滤波器的参数如下:ζ=0.7,ω
n=500。
选择饱和非线性输入为:
仿真是在系统参数和非线性函数未知的前提下进行的。对于考虑铁损的自适应神经网络控制方法的仿真结果如附图所示。跟踪信号和期望信号如图2所示,位置跟踪误差如图3所示。由图2和图3看出,系统的输出可以很好的跟踪期望信号;d轴定子电压和q轴定子电压如图4和图5所示,由图4和图5看出,控制器输入u
d和u
q都稳定在一个有界区域内。
以上模拟信号清楚地表明,本发明中考虑铁损的永磁同步电机随机命令滤波神经网络控制方法可以高效地跟踪参考信号,因此,具有良好实际实施意义。
当然,以上说明仅仅为本发明的较佳实施例,本发明并不限于列举上述实施例,应当说明的是,任何熟悉本领域的技术人员在本说明书的教导下,所做出的所有等同替代、明显变形形式,均落在本说明书的实质范围之内,理应受到本发明的保护。
Claims (1)
- 考虑铁损的永磁同步电机随机命令滤波神经网络控制方法,其特征在于,包括如下步骤:a建立考虑铁损的永磁同步电机的d-q坐标轴动态数学模型,如公式(1)所示:其中,u d,u q表示定子电压;Θ和ω分别表示转子角位置和转子角速度;i d,i q表示定子电流;i od,i oq表示励磁电流;J表示转动惯量;n p表示极对数;L d,L q表示定子电感;L ld,L lq表示定子漏感;L md,L mq表示励磁电感;R 1表示定子电阻,R c表示铁芯损耗电阻;T L表示负载转矩;λ PM表示转子永磁体产生的磁通量;为了简化计算过程,定义新的变量如下:则考虑铁损的永磁同步电机的动态数学模型可表示为:其中,ψ 2,ψ 3,ψ 4,ψ 5,ψ 6均表示未知的光滑非线性函数;b基于Lyapunov函数,对于任意给定的函数V=V(x)∈C 2,C 2表示复数集;定义差分运算,由伊藤微分法则可知:对于公式(1),u为永磁同步电机系统的对称饱和非线性输入;u可描述为:其中,u max>0和u min<0是未知的输入饱和常数;v是饱和非线性的输入信号;通过使用分段的平滑函数来近似饱和函数,定义如下函数:将公式(3)中的sat(v)写成:u=sat(v)=g(v)+d(v);其中,d(v)=sat(v)-g(v)是有界函数,其边界可表示为:|d(v)|=|sat(v)-g(v)|≤max{u max(1-tanh(1)),u min(tanh(1)-1)}=D (4)其中,D为常数;存在常数b m和b M,使0<b m≤g ia≤b M<∞成立;其中,g ia为x ia+1的系数,ia=1,…,n;假设存在一个C 2的函数V(x):R n→R +,取两个常数e 0>0和g 0>0,则存在:其中,R n表示全体n维实数列向量构成的集合,R +表示正数;其中,E[V(x)]表示函数V(x)的期望,V(x 0)表示当x=x 0时V(x)的初始值;命令滤波器定义如下:其中,φ 1和φ 2表示实数,ω n>0,ζ∈(0,1];其中,ε>0,p>1,q>1,并且(p-1)(q-1)=1;其中,δ(Z)为跟踪误差,且满足|δ(Z)|≤ε;当N的取值为N *时,|δ(Z)|达到最小;下面给出基于反步法的随机非线性神经网络控制器的设计:定义系统误差变量:上式中,x 1d为跟踪信号,x 1,c,x 2,c,x 3,c,x 5,c分别为输入信号α 1,α 2,α 3,α 5时滤波器的输出信号,ξ i为误差补偿信号,i=1,2,3,4,5,6;定义未知常数θ j=||N j|| 2,j=2,3,4,5,6,N j表示未知理想权向量, 其中, 表示θ j的估计值;P i(Z)=P i表示基函数向量,δ i(Z)为跟踪误差,且满足|δ i(Z) |≤ε i;控制方法设计的每一步都会选取一个Lyapunov函数来构建一个虚拟控制函数或者真实控制律,控制方法的设计具体包括以下步骤:b1定义误差信号z 1=x 1-x 1d,补偿误差v 1=z 1-ξ 1;其中,k 1表示正设计参数;b2定义误差信号z 2=x 2-x 1,c,补偿误差v 2=z 2-ξ 2;根据杨氏不等式可得:其中,d表示|T L|的上限值;d>0;b3定义误差信号z 3=x 3-x 2,c,补偿误差v 3=z 3-ξ 3;由杨氏不等式可得:其中,l 3为选取的正常数,由此可得:定义x 4=v 4+ξ 4+x 3,c;b4定义误差信号z 4=x 4-x 3,c,补偿误差v 4=z 4-ξ 4;其中,l 4为选取的正常数,将该公式代入公式(21)可得:式中,k 4表示正设计参数,d q为正数;由杨氏不等式得到:其中,D q为常数;b5定义误差信号z 5=x 5,补偿误差v 5=z 5-ξ 5;其中,l 5为选取的正常数,可得:其中,k 5表示正设计参数,可得到:b6定义误差信号z 6=x 6-x 5,c,补偿误差v 6=z 6-ξ 6;由杨氏不等式得:其中,d d表示常数;根据上述公式可知:其中,D d表示常数,d 2为常数;通过上述公式可得:c由上述分析,选择Lyapunov函数如下:其中,p j和λ j均表示设计常数;取a 0,b 0分别为:a 0=min{4e 1,4e 2,4e 3,4e 4,4e 5,4e 6,p 2,p 3,p 4,p 5,p 6};因此,公式(36)可表示为:LV≤-a 0V+b 0,t≥0;式(36)可表示为:上式可改写为:E[|v i| 4]≤4E[V(t)],i=1,2,3,4,5,6 (38)构造Lyapunov函数V ξ如下:定义滤波误差为(x t,c-α t)且满足|x t,c-α t|≤σ t,t=1,2,3,5,由式(39)可得:
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910084184.7A CN109921698B (zh) | 2019-01-29 | 2019-01-29 | 考虑铁损的永磁同步电机随机命令滤波神经网络控制方法 |
CN201910084184.7 | 2019-01-29 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020155509A1 true WO2020155509A1 (zh) | 2020-08-06 |
Family
ID=66961071
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/090499 WO2020155509A1 (zh) | 2019-01-29 | 2019-06-10 | 考虑铁损的永磁同步电机随机命令滤波神经网络控制方法 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN109921698B (zh) |
WO (1) | WO2020155509A1 (zh) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110336505B (zh) * | 2019-07-10 | 2020-08-25 | 青岛大学 | 基于状态约束的异步电动机命令滤波模糊控制方法 |
CN110401390B (zh) * | 2019-07-22 | 2021-02-26 | 青岛大学 | 基于观测器的永磁同步电机随机命令滤波模糊控制方法 |
CN111200381B (zh) * | 2020-01-03 | 2023-08-22 | 江苏大学 | 一种新能源汽车驱动电机鲁棒最优抗干扰控制器的构造方法 |
CN112039374B (zh) * | 2020-08-05 | 2022-04-08 | 青岛大学 | 考虑输入饱和的永磁同步电动机命令滤波离散控制方法 |
CN113659895B (zh) * | 2021-08-02 | 2023-08-11 | 青岛大学 | 基于指令滤波的永磁同步电动机全状态约束有限时间控制方法 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104993760A (zh) * | 2015-07-31 | 2015-10-21 | 青岛大学 | 考虑铁损的电动汽车异步电机模糊自适应动态面控制方法 |
CN106533313A (zh) * | 2016-11-28 | 2017-03-22 | 青岛大学 | 电动汽车用永磁同步电机的极限学习机命令滤波控制方法 |
CN106655962A (zh) * | 2016-11-28 | 2017-05-10 | 青岛大学 | 基于极限学习机的电动汽车异步电机驱动系统控制方法 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9754204B2 (en) * | 2013-08-05 | 2017-09-05 | Board Of Trustees Of The University Of Alabama | Systems, methods and devices for vector control of permanent magnet synchronous machines using artificial neural networks |
CN104467601A (zh) * | 2014-12-19 | 2015-03-25 | 东南大学 | 一种基于wams的同步发电机参数辨识及控制方法 |
CN108964545B (zh) * | 2018-07-30 | 2019-11-19 | 青岛大学 | 一种基于命令滤波的同步电机神经网络反步离散控制方法 |
-
2019
- 2019-01-29 CN CN201910084184.7A patent/CN109921698B/zh active Active
- 2019-06-10 WO PCT/CN2019/090499 patent/WO2020155509A1/zh active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104993760A (zh) * | 2015-07-31 | 2015-10-21 | 青岛大学 | 考虑铁损的电动汽车异步电机模糊自适应动态面控制方法 |
CN106533313A (zh) * | 2016-11-28 | 2017-03-22 | 青岛大学 | 电动汽车用永磁同步电机的极限学习机命令滤波控制方法 |
CN106655962A (zh) * | 2016-11-28 | 2017-05-10 | 青岛大学 | 基于极限学习机的电动汽车异步电机驱动系统控制方法 |
Also Published As
Publication number | Publication date |
---|---|
CN109921698A (zh) | 2019-06-21 |
CN109921698B (zh) | 2020-06-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020155509A1 (zh) | 考虑铁损的永磁同步电机随机命令滤波神经网络控制方法 | |
CN106788086B (zh) | 考虑输入饱和的异步电机命令滤波有限时间模糊控制方法 | |
CN110401391B (zh) | 异步电动机随机系统模糊自适应动态面控制方法 | |
CN106788046B (zh) | 永磁同步电机命令滤波有限时间模糊控制方法 | |
CN110401390B (zh) | 基于观测器的永磁同步电机随机命令滤波模糊控制方法 | |
CN109873582B (zh) | 基于动态面的永磁同步电机有限时间位置跟踪控制方法 | |
CN107240921A (zh) | 基于积分自适应逆推的svc滑模控制方法 | |
CN110112971B (zh) | 一种基于有限时间动态面的异步电动机位置跟踪控制方法 | |
CN108322117B (zh) | 无刷双馈电机的控制系统、前馈控制方法及参数辨识方法 | |
Tir et al. | Adaptive high gain observer based MRAS for sensorless induction motor drives | |
Fereka et al. | MRAS sensorless speed control of an induction motor drive based on fuzzy sliding mode control | |
CN115313939A (zh) | 永磁同步电机随机系统指令滤波模糊自适应控制方法 | |
CN106788053B (zh) | 基于观测器的电动车永磁同步电机系统误差补偿控制方法 | |
CN109873583B (zh) | 基于状态受限的永磁同步电机模糊位置跟踪控制方法 | |
CN114944801A (zh) | 一种基于新息自适应扩展卡尔曼的pmsm无位置传感器控制方法 | |
CN117895851A (zh) | 一种表贴式永磁同步电机全速域控制方法 | |
Zhang et al. | Observer-based fuzzy adaptive formation control for saturated MIMO nonlinear multiagent systems under switched topologies | |
CN113659895A (zh) | 基于指令滤波的永磁同步电动机全状态约束有限时间控制方法 | |
CN106533313B (zh) | 电动汽车用永磁同步电机的极限学习机命令滤波控制方法 | |
Tsai et al. | Implementation of a DSP‐based speed‐sensorless adaptive control for permanent‐magnet synchronous motor drives with uncertain parameters using linear matrix inequality approach | |
CN113381662B (zh) | 基于有限时间动态面技术的永磁同步电动机随机系统模糊控制方法 | |
Wei et al. | Model-Free Predictive Control Using Sinusoidal Generalized Universal Model for PMSM Drives | |
CN111293941B (zh) | 一种考虑铁损的永磁同步电动机有限时间动态面控制方法 | |
Mohamed et al. | New hybrid sensorless speed of a non-salient pole PMSG coupled to wind turbine using a modified switching algorithm | |
CN112019111B (zh) | 基于状态约束的永磁同步电动机随机系统模糊自适应控制方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19912238 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19912238 Country of ref document: EP Kind code of ref document: A1 |