符号定义
- 与参考文献2中一致
Popov超稳定性概述1
对于连续时间线性定常系统,超稳定性成立的条件有两个:
- 输入输出积分满足Popov积分不等式:
∫ 0 T u T ( t ) y ( t ) d t ≤ δ ( ∣ ∣ x ( 0 ) ∣ ∣ ) s u p 0 ≤ t ≤ T ∣ ∣ x ( t ) ∣ ∣ \int_{0}^{T}u^T(t)y(t)dt\leq\delta(||x(0)||)sup_{0\leq t \leq T}||x(t)|| ∫0TuT(t)y(t)dt≤δ(∣∣x(0)∣∣)sup0≤t≤T∣∣x(t)∣∣ - 传递函数矩阵满足正实性。
在MRAS中的应用,以PMSM参数辨识为例
在参考文献2中,可以看到本身模型参考自适应原理比较简单,只是对于控制器设计和稳定性证明比较麻烦,用到了Popov超稳定性理论来设计控制器。
首先,参考模型选择与源模型相同,构造了一误差系统,只要保证该误差系统的状态变量收敛到0,则电机参数即可估计出来。
e
˙
=
(
A
+
G
)
e
+
Δ
A
i
^
+
Δ
B
u
+
Δ
C
\dot{e}=(A+G)e+\Delta A\hat{i}+\Delta Bu+\Delta C
e˙=(A+G)e+ΔAi^+ΔBu+ΔC
其中
e
e
e为误差矢量,
e
=
i
−
i
^
,
i
=
[
i
d
i
q
]
T
,
Δ
A
=
A
−
A
^
,
Δ
B
=
B
−
B
^
,
Δ
C
=
C
−
C
^
e = i - \hat{i}, i = \left[ \begin{matrix} i_d & i_q \end{matrix} \right]^T, \Delta A = A - \hat{A}, \Delta B = B - \hat{B}, \Delta C = C - \hat{C}
e=i−i^,i=[idiq]T,ΔA=A−A^,ΔB=B−B^,ΔC=C−C^。取
w
=
−
(
Δ
A
i
^
+
Δ
B
u
+
Δ
C
)
w = -(\Delta A\hat{i}+\Delta Bu+\Delta C)
w=−(ΔAi^+ΔBu+ΔC),则有:
e
˙
=
(
A
+
G
)
e
−
w
\dot{e}=(A+G)e-w
e˙=(A+G)e−w
通过设计
G
G
G来保证系统传递函数矩阵严格正实3,设计
−
w
-w
−w来保证满足输入输出Popov积分不等式。在该系统中,有
A
=
[
−
a
w
e
−
w
e
−
a
]
,
A
^
=
[
−
a
^
w
e
−
w
e
−
a
^
]
,
B
=
[
b
0
0
b
]
,
B
^
=
[
b
^
0
0
b
^
]
,
C
=
[
0
−
w
e
c
]
T
,
C
^
=
[
0
−
w
e
c
^
]
T
A = \left[ \begin{matrix} -a & w_e\\ -w_e & -a \end{matrix} \right], \hat{A} = \left[ \begin{matrix} -\hat{a} & w_e\\ -w_e & -\hat{a} \end{matrix} \right], B = \left[ \begin{matrix} b & 0\\ 0 & b \end{matrix} \right], \hat{B} = \left[ \begin{matrix} \hat{b} & 0\\ 0 & \hat{b} \end{matrix} \right], C = \left[ \begin{matrix} 0 & -w_ec \end{matrix} \right]^T, \hat{C} = \left[ \begin{matrix} 0 & -w_e\hat{c} \end{matrix} \right]^T
A=[−a−wewe−a],A^=[−a^−wewe−a^],B=[b00b],B^=[b^00b^],C=[0−wec]T,C^=[0−wec^]T,取
a
^
,
b
^
,
c
^
\hat{a}, \hat{b}, \hat{c}
a^,b^,c^均为PI类型的控制器,有:
a
^
=
∫
0
t
f
1
(
τ
)
d
τ
+
f
2
(
t
)
+
a
^
(
0
)
b
^
=
∫
0
t
g
1
(
τ
)
d
τ
+
g
2
(
t
)
+
b
^
(
0
)
c
^
=
∫
0
t
h
1
(
τ
)
d
τ
+
h
2
(
t
)
+
c
^
(
0
)
\hat{a}=\int_{0}^{t}f_1(\tau)d\tau + f_2(t) + \hat{a}(0) \\ \hat{b}=\int_{0}^{t}g_1(\tau)d\tau + g_2(t) + \hat{b}(0) \\ \hat{c}=\int_{0}^{t}h_1(\tau)d\tau + h_2(t) + \hat{c}(0)
a^=∫0tf1(τ)dτ+f2(t)+a^(0)b^=∫0tg1(τ)dτ+g2(t)+b^(0)c^=∫0th1(τ)dτ+h2(t)+c^(0)
将上
a
^
,
b
^
,
c
^
\hat{a},\hat{b},\hat{c}
a^,b^,c^计算式带入到Popov积分不等式中,其中
u
(
t
)
u(t)
u(t)即为误差系统中的
−
w
-w
−w,
y
(
t
)
y(t)
y(t)则为误差系统中的
e
e
e,借助matlab符号运算,即可得到化简后的Popov积分不等式如下:
∫
0
T
e
q
w
e
(
c
−
c
^
)
+
(
e
q
i
q
^
+
e
d
i
d
^
)
(
a
−
a
^
)
−
(
e
q
u
q
+
e
d
u
d
)
(
b
−
b
^
)
d
t
≥
−
γ
0
2
⇒
∫
0
T
e
q
w
e
(
c
−
(
∫
0
t
h
1
(
τ
)
d
τ
+
h
2
(
t
)
+
c
^
(
0
)
)
)
d
t
+
∫
0
T
(
e
q
i
q
^
+
e
d
i
d
^
)
(
a
−
(
∫
0
t
f
1
(
τ
)
d
τ
+
f
2
(
t
)
+
a
^
(
0
)
)
)
d
t
−
∫
0
T
(
e
q
u
q
+
e
d
u
d
)
(
b
−
(
∫
0
t
g
1
(
τ
)
d
τ
+
g
2
(
t
)
+
b
^
(
0
)
)
)
d
t
≥
−
γ
0
2
\int_{0}^{T}e_qw_e(c-\hat{c}) + (e_q\hat{i_q}+e_d\hat{i_d})(a-\hat{a}) - (e_qu_q + e_du_d)(b-\hat{b})dt \geq -\gamma_0^2 \\ \Rightarrow \\ \int_{0}^{T}e_qw_e(c-(\int_{0}^{t}h_1(\tau)d\tau + h_2(t) + \hat{c}(0)))dt \\ + \int_{0}^{T}(e_q\hat{i_q}+e_d\hat{i_d})(a-(\int_{0}^{t}f_1(\tau)d\tau + f_2(t) + \hat{a}(0)))dt \\ - \int_{0}^{T}(e_qu_q + e_du_d)(b-(\int_{0}^{t}g_1(\tau)d\tau + g_2(t) + \hat{b}(0)))dt \\ \geq -\gamma_0^2
∫0Teqwe(c−c^)+(eqiq^+edid^)(a−a^)−(equq+edud)(b−b^)dt≥−γ02⇒∫0Teqwe(c−(∫0th1(τ)dτ+h2(t)+c^(0)))dt+∫0T(eqiq^+edid^)(a−(∫0tf1(τ)dτ+f2(t)+a^(0)))dt−∫0T(equq+edud)(b−(∫0tg1(τ)dτ+g2(t)+b^(0)))dt≥−γ02
其中,
γ
0
\gamma_0
γ0为一误差系统中与系统变量(即误差
e
e
e)初值相关的量。
e
d
=
i
d
−
i
d
^
,
e
q
=
i
q
−
i
q
^
e_d=i_d-\hat{i_d},e_q=i_q-\hat{i_q}
ed=id−id^,eq=iq−iq^可以看到,要满足上式,即使:
∫
0
T
e
q
w
e
(
c
−
(
∫
0
t
h
1
(
τ
)
d
τ
+
h
2
(
t
)
+
c
^
(
0
)
)
)
d
t
≥
−
γ
1
2
∫
0
T
(
e
q
i
q
^
+
e
d
i
d
^
)
(
a
−
(
∫
0
t
f
1
(
τ
)
d
τ
+
f
2
(
t
)
+
a
^
(
0
)
)
)
d
t
≥
−
γ
2
2
∫
0
T
(
e
q
u
q
+
e
d
u
d
)
(
b
−
(
∫
0
t
g
1
(
τ
)
d
τ
+
g
2
(
t
)
+
b
^
(
0
)
)
)
d
t
≥
−
γ
3
2
\int_{0}^{T}e_qw_e(c-(\int_{0}^{t}h_1(\tau)d\tau + h_2(t) + \hat{c}(0)))dt \geq -\gamma_1^2 \\ \int_{0}^{T}(e_q\hat{i_q}+e_d\hat{i_d})(a-(\int_{0}^{t}f_1(\tau)d\tau + f_2(t) + \hat{a}(0)))dt \geq -\gamma_2^2 \\ \int_{0}^{T}(e_qu_q + e_du_d)(b-(\int_{0}^{t}g_1(\tau)d\tau + g_2(t) + \hat{b}(0)))dt \geq -\gamma_3^2
∫0Teqwe(c−(∫0th1(τ)dτ+h2(t)+c^(0)))dt≥−γ12∫0T(eqiq^+edid^)(a−(∫0tf1(τ)dτ+f2(t)+a^(0)))dt≥−γ22∫0T(equq+edud)(b−(∫0tg1(τ)dτ+g2(t)+b^(0)))dt≥−γ32
均满足即可。以上三式中第一个不等式为例,将其拆开,可以得到:
∫
0
T
e
q
w
e
(
c
−
∫
0
t
h
1
(
τ
)
d
τ
−
c
^
(
0
)
)
d
t
≥
−
γ
11
2
∫
0
T
−
e
q
w
e
h
2
(
t
)
d
t
≥
−
γ
12
2
\int_{0}^{T}e_qw_e(c-\int_{0}^{t}h_1(\tau)d\tau - \hat{c}(0))dt \geq -\gamma_{11}^2 \\ \int_{0}^{T}-e_qw_eh_2(t)dt \geq -\gamma_{12}^2
∫0Teqwe(c−∫0th1(τ)dτ−c^(0))dt≥−γ112∫0T−eqweh2(t)dt≥−γ122
均满足即可。对于上两式中的第一式,可以利用如下不等式:
∫
0
T
d
f
(
t
)
d
t
k
f
(
t
)
d
t
=
k
2
[
f
2
(
T
)
−
f
2
(
0
)
]
≥
k
2
f
2
(
0
)
\int_{0}^{T}\frac{df(t)}{dt}kf(t)dt=\frac{k}{2}[f^2(T)-f^2(0)] \geq \frac{k}{2}f^2(0)
∫0Tdtdf(t)kf(t)dt=2k[f2(T)−f2(0)]≥2kf2(0)
取
d
f
(
t
)
d
t
=
e
q
w
e
,
k
f
(
t
)
=
c
−
∫
0
t
h
1
(
τ
)
d
τ
−
c
^
(
0
)
\frac{df(t)}{dt} =e_qw_e, kf(t)=c-\int_{0}^{t}h_1(\tau)d\tau - \hat{c}(0)
dtdf(t)=eqwe,kf(t)=c−∫0th1(τ)dτ−c^(0),则可以得到:
h
1
(
t
)
=
−
e
q
w
e
K
h
i
,
K
h
i
≥
0
h_1(t)=-e_qw_eK_{hi},K_{hi} \geq 0
h1(t)=−eqweKhi,Khi≥0
而对于上两式中的第二式,可以直接取
h
2
(
t
)
=
−
K
h
p
e
q
w
e
h_2(t)=-K_{hp}e_qw_e
h2(t)=−Khpeqwe即可保证不等式成立:
h
2
(
t
)
=
−
K
h
p
e
q
w
e
,
K
h
p
≥
0
h_2(t)=-K_{hp}e_qw_e,K_{hp} \geq 0
h2(t)=−Khpeqwe,Khp≥0
因此,对于
c
^
\hat{c}
c^的控制率可以选择:
c
^
=
−
K
h
i
∫
0
t
e
q
(
τ
)
w
e
d
τ
−
K
h
p
e
q
(
t
)
w
e
+
c
^
(
0
)
\hat{c}=-K_{hi}\int_{0}^{t}e_q(\tau)w_ed\tau - K_{hp}e_q(t)w_e + \hat{c}(0)
c^=−Khi∫0teq(τ)wedτ−Khpeq(t)we+c^(0)
即可保证误差系统满足Popov超稳定性条件。使用同样的方法,即可得到
a
^
与
b
^
\hat{a}与\hat{b}
a^与b^的控制率如下:
a
^
=
−
K
f
i
∫
0
t
(
i
d
^
(
τ
)
e
d
(
τ
)
+
i
q
^
(
τ
)
e
q
(
τ
)
)
d
τ
−
K
f
p
(
i
d
^
(
τ
)
e
d
(
τ
)
+
i
q
^
(
τ
)
e
q
(
τ
)
)
+
a
^
(
0
)
b
^
=
K
g
i
∫
0
t
(
u
d
(
τ
)
e
d
(
τ
)
+
u
q
(
τ
)
e
q
(
τ
)
)
d
τ
+
K
g
p
(
u
d
(
t
)
e
d
(
t
)
+
u
q
(
t
)
e
q
(
t
)
)
+
b
^
(
0
)
\hat{a}=-K_{fi}\int_{0}^{t}(\hat{i_d}(\tau)e_d(\tau)+\hat{i_q}(\tau)e_q(\tau))d\tau - K_{fp}(\hat{i_d}(\tau)e_d(\tau)+\hat{i_q}(\tau)e_q(\tau)) + \hat{a}(0) \\ \hat{b}=K_{gi}\int_{0}^{t}(u_d(\tau)e_d(\tau)+u_q(\tau)e_q(\tau))d\tau + K_{gp}(u_d(t)e_d(t)+u_q(t)e_q(t)) + \hat{b}(0)
a^=−Kfi∫0t(id^(τ)ed(τ)+iq^(τ)eq(τ))dτ−Kfp(id^(τ)ed(τ)+iq^(τ)eq(τ))+a^(0)b^=Kgi∫0t(ud(τ)ed(τ)+uq(τ)eq(τ))dτ+Kgp(ud(t)ed(t)+uq(t)eq(t))+b^(0)
Note
值得注意的是,文献2中利用MRAS同时辨识出三个电机参数,但实际上系统模型的阶数仅为两阶,因此个人觉得应该是有些许错误,在实际仿真时也印证了这一点:只有两个参数时辨识才准确,若三个参数同时辨识,结果将不准确。
参考文献
- [2] Quntao An and Li Sun, “On-line parameter identification for vector controlled PMSM drives using adaptive algorithm,” 2008 IEEE Vehicle Power and Propulsion Conference, Harbin, 2008, pp. 1-6, doi: 10.1109/VPPC.2008.4677634.
- [3] Xu Junfeng, Xu Yinglei, Xu jiangping, et al. “A new control method for permanent magnet synchronous machines with observer”, Aachen Germany: 35th IEEE Power Electronics Specialists Conference, 2004.
附录
- matlab公式化简源码
clc;
clear all;
syms a b c A B C Ag Bg Cg e dltA dltB dltC i ig we id iq idg iqg ag bg cg e ud uq u real
A = [-a we;-we -a]
Ag = [-ag we;-we -ag]
B = [b 0;0 b]
Bg = [bg 0;0 bg]
C = [0;-we*c]
Cg = [0;-we*cg]
dltA = A - Ag
dltB = B - Bg
dltC = C - Cg
i = [id;iq]
ig = [idg;iqg]
e = i - ig
u = [ud;uq]
clc
-(dltA*ig + dltB*u + dltC)' * e