Professional Documents
Culture Documents
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI
ABSTRACT This paper is concerned with velocity control in a permanent magnet synchronous motor
(PMSM) when it is fed by an inverter-DC/DC Buck power converter system as power amplifier. We present,
for the first time, a formal local asymptotic stability proof to solve this control problem. We stress that
this is the first time that this problem is solved for an AC motor. Our control scheme is simple when
compared to differential flatness- and backsteping-based proposals in the literature to solve this problem
for DC motors. The key for these achievements is the employment of a novel passivity-based approach
which takes advantage of the natural energy exchange among the electrical and mechanical subsystems
that compose the inverter-DC/DC Buck power converter-PMSM system. The main features of this novel
passivity-based approach are summarized in this paper.
INDEX TERMS Energy-based control, inverter-dc/dc buck power converter system, Lyapunov stability,
permanent magnet synchronous motors, velocity control.
VOLUME 4, 2016 1
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
that AC motor models, and in particular the PMSM model, times to have a three-phase system that feed the three phases
are well known to be nonlinear and multi input-multi output. of the PMSM. The symbols icj , Vj , Ij , j = 1, 2, 3, represent
This renders much more complex the control design task and the electric current through the inductance L, voltage at the
represents the merit of our contribution. capacitor C terminals, and electric current through the j−th
The features of our proposal are the following. phase of the PMSM. Because of the transient responses, the
• The proposed control scheme is composed by four three-phase source voltages are not ensured to be balanced,
PI control loops. This results in a simple and robust i.e. V1 (t) + V2 (t) + V3 (t) = 0 is not true for all the time.
control law as opposite to control laws obtained when The balanced voltages assumption is a standard argument to
applying the control techniques (i.e. differential flatness obtain the standard dq dynamical model of a PMSM having
and backstepping) that have been proposed to solve this an isolated neutral (see [21], Ch. 7). Hence, we also show
problem in DC-motors. It is the authors belief that this how to handle the unbalanced voltages assumption to retrieve
is the very reason why any work solving this problem the standard dq dynamical model of PMSM’s. The symbol E
for AC motors has not been reported until now. stands for voltage of the DC power supply. The system inputs
• Our proposal relies on a novel passivity-based approach are vj which only take the discrete values {+1, −1, 0} repre-
which exploits the energy exchange that naturally exists senting the on-positive, the on-negative and the discharging
among the electrical and mechanical subsystems that states of transistors Q1j , Q2j , Q̄1j , Q̄2j [18].
compose the inverter-DC/DC Buck power electronic
converter-PMSM system. This feature is instrumental Q 11 Q 21
to design a simple control law because such energy + v1 à v1 I1
ic 1 +
exchange is represented by the natural cancellation of
V1 Rc
many terms in the stability analysis. If this was not Q 11 Q 21
1
à
the case such large amount of terms should be exactly + v1 à v1
cancelled by computing and feeding back them. PMSM
• Contrary to standard passivity-based approaches as that Q 12 Q 22
ïn
ê
+ v2 I2
in [19], and other more recent approaches as that in [20] à v2
ic 2 + 2
which might be tried to solve this problem, our approach 3
V2 Rc
does not require to feedback the time derivative of either Q 12 Q 22
à
the desired electric current nor the desired voltages. This + v2 à v2 ïn
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
Using the dq transformation [22]: is the motor inductance matrix and Γ is the flux linkages
due to the motor permanent magnet. Then applying the dq
x = T T xN , xN = [xq , xd , x0 ]T , (5) transformation defined in (5) to (8) and using (7) we find (see
r cos(θ) cos θ − 2π 2π
3 cos θ + 3 [22] for further details):
2
T = sin(θ) sin θ − 2π 3 sin θ + 2π
3 ,
(6)
3 √1 √1 √1 Lq I˙q = −RIq − np Ld Id ω − ΦM ω + Vq , (9)
2 2 2
Ld I˙d = −RId + np Lq Iq ω + Vd , (10)
where x stands for U, V, I, Ic , and θ = np q with q the
mechanical rotor position and np the motor number of pole L0 I˙0 = −RI0 , (11)
pairs, we have that (3) and (4) can be written as (recall that J ω̇ = −bω + np (Ld − Lq )Id Iq + ΦM Iq − τL ,(12)
T T = T −1 ): where the positive constant scalars Ld , Lq , L0 stand for the
LI˙cN = −np LG1 IcN ω − VN + EUN , dq0 phase inductances, ω = q̇ is the motor velocity, ΦM , b, J,
1 are positive constants standing for torque constant, viscous
C V̇N = −np CG1 VN ω + IcN − IN − VN ,
R c q τL is
friction coefficient and rotor inertia, respectively, and
load torque. Notice that, according to (5), (6), I0 = 13 (I1 +
0 1 0 q
G1 = −1 0 0 . I2 + I3 ) and V0 = 1
3 (V1 + V2 + V3 ). Moreover, the star
0 0 0 connection with isolated neutral of the motor phase windings
ensures that I0 = 0 for all time. This is consistent with the
fact that (11) has no input and it is exponentially stable. We
1
ï
1
also stress that in the case of balanced source voltages (11)
I1
converts into L0 I˙0 = −RI0 +V0 , which is consistent with the
+ V1 balanced current condition, i.e. I0 = 0, since V0 = 0 when
10 the source voltages are balanced. Thus, (11) is also consistent
n n
ê with the unbalanced source voltages condition because V0 6=
V3 V2 20
30 3 0 is not present, ensuring I0 = 0 for all time.
+
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
The terms C(Vd2 + Vq2 )/2 stand for electric energy stored in (18) in closed-loop with the following controller:
2
capacitors of the Buck power converters, whereas L(Icd + Z t
2 1 ∗ ˜ ˜
Icq )/2 represent the magnetic energy stored in inductances Uq = Vq − Kpcq Icq − Kicq Icq dr , (20)
E
of the Buck power converters, and (Lq Iq2 + Ld Id2 )/2 stand 0
Z t
for the magnetic energy stored in the electrical subsystem of 1
Ud = Vd∗ − Kpcd I˜cd − Kicd I˜cd dr , (21)
the PMSM. Finally, 12 Jω 2 is the kinetic energy stored in the E 0
mechanical subsystem of the PMSM. The time derivative of U0 = 0, (22)
V along the trajectories of system in (9)-(18) is given as: ∗ 1 ∗ ∗
Icq = V − KpV q Ṽq + Iq (23)
Rc q
Z t
LKpV q ˜
−KiV q Ṽq + np Lω I˜cd + Icq dr,
0 C
1
V̇ = Vd [np CVq ω + Icd − Id − Vd ] ∗ 1 ∗
Rc Icd = V − KpV d Ṽd + Id∗ (24)
Rc d
1 Z t
+Vq [−np CVd ω + Icq − Iq −
Vq ] LKpV d ˜
Rc −KiV d Ṽd − np Lω I˜cq + Icd dr,
+Icd [np LIcq ω − Vd + EUd ] 0 C
Z t
1
+Icq [−np LIcd ω − Vq + EUq ] Id∗ = 0, Iq∗ = (−kp ω̃ − ki σ(z)), z = ω̃dr,(25)
ΦM 0
+Iq [−RIq − np Ld Id ω − ΦM ω + Vq ] Z t
np L ˜
+Id [−RId + np Lq Iq ω + Vd ] Vd∗ = −αpd Id − αid Id − np Cω Ṽq − Icq ω
0 Rc
+ω[−bω + np (Ld − Lq )Id Iq + ΦM Iq − τL ]. Lαpd
1
+ + KpV d I˜cd
Ld Rc
Cαpd
−Lnp KpV q ω I˜cq + Ṽd dr, (26)
Notice that several terms cancel to obtain: Ld
Z t
np L ˜
Vq∗ = −αpq I˜q − αiq I˜q + np Cω Ṽd + ω Icd
0 Rc
Lαpq 1
1 + + KpV q I˜cq
V̇ = − (V 2 + Vd2 ) − R(Iq2 + Id2 ) − bω 2 − ωτL Lq Rc
Rc q
˜ Cαpq
+EIcd Ud + EIcq Uq . +Lnp KpV d ω Icd + Ṽq dr, (27)
Lq
where:
I˜ci ∗
= Ici − Ici , Ṽi = Vi − Vi∗ , (28)
We stress that these term cancellations represent 1) energy
exchange between the electrical and the mechanical sub- ω̃ = ω − ω ∗ , I˜i = Ii − Ii∗ , (29)
systems of the PMSM, 2) energy exchange between the with subindex i standing for d and q and Id∗ = 0, ω ∗ is a real
capacitor and the electrical subsystem of the PMSM, 3) constant standing for the desired velocity, σ(z) is a strictly
energy exchange between the capacitor and the inductance increasing linear saturation function for some (L∗ , M ) (see
of the Buck power electronic converter, 4) energy exchange Definition 1). Furthermore, it is also required that function
between the dq phases of the inductor of the Buck power σ(z) be continuously differentiable such that:
electronic converter, and 5) energy exchange between the
dq phases of the capacitor of the Buck power electronic dσ(z)
0< ≤ 1, ∀z ∈ R. (30)
converter. dz
There always exist positive constants kp , ki , αpd , αpq , KpV q ,
Hence, if we define the input [EUq , EUd , −τL ]T and the KpV d , Kpcq , Kpcd , Kicq , Kicd , KiV q , KiV d , αid , αiq , β, such
output [Icq , Icd , ω]T , then the dynamical model in (9)-(18) that the origin of the closed-loop system is locally asymptot-
is passive. These properties are exploited in this paper to ically stable.
design a velocity controller for the inverter-DC/DC Buck
power converter-PMSM system. A. CLOSED-LOOP DYNAMICS
First notice that the zero sequence dynamics (11), (15), (18),
is globally exponentially stable. Also notice that the remain-
IV. MAIN RESULT ing expressions in the model (9)-(18) are independent from
the zero sequence variables, i.e. Ic0 , V0 , I0 . Thus, choosing
Our main result is stated in the following proposition.
U0 = 0 in (22) is enough to ensure that I0 = 0 and both
Proposition 1: Consider the mathematical model in (9)- Ic0 → 0, V0 → 0 exponentially as time grows. This means
4 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
6 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
These observations motivate the use of the following “en- is given as:
ergy” storage function for the closed-loop dynamics (see
1
(19)): Ẇ = I˜cd np L (−αpq I˜q ) − KpV q Ṽq
Rc
1
+ (−kp ω̃ − ki χ(z)) + ξ1 ω̃
ΦM
C 2 L 2 +np LI˜cq ω ∗ − Kpcd I˜cd − LI˙cd
∗
W (y) = (Ṽ + Ṽq2 ) + (I˜cd + I˜cq
2
)
2 d 2
1
1 1 +np L (−αpq I˜q ) − KpV q Ṽq
+ (Lq I˜q2 + Ld Id2 ) + (αiq zq2 + αid zd2 ) (63) Rc
2 2
1 1
+ (KiV d ζvd2 2
+ KiV q ζvq ) + (−kp ω̃ − ki χ(z)) ω ∗
2 Φ
M
1 τL + bω ∗
1
+ (Kicd ζd2 + Kicq ζq2 ) + Vω (ω̃, z + ), ˜
+Icq − np L (−αpd Id ) − KpV d Ṽd + ξ2 ω̃
2 ki Rc
| {z }
:=κ
˜ ∗ 1
−np LIcd ω − np L (−αpd Id )
Rc
−KpV d Ṽd ω ∗ − Kpcq I˜cq − LI˙cq ∗
where:
+Ṽd np C(−αpq I˜q + ξ)ω̃ + np C(−αpq I˜q )ω ∗
τL + bω ∗ 1 1 ∗
Vω (ω̃, z + ) = J ω̃ 2 − + KpV d Ṽd − C V̇d
ki 2 Rc
Z z
+[ki + β(b + kp )] χ(r)dr + βJχ(z)ω̃. +Ṽq − np C(−αpd Id − δ)ω̃ − np C(−αpd Id )ω ∗
∗ τL +bω
− ki
1
− + KpV q Ṽq − C V̇q∗
Rc
∗
+Id [−(R + αpd )Id + np Lq I˜q ω ∗
We stress that function Vω (ω̃, z + τL +bω
ki ), defined in (63), np Lq ki ∗ np Lq kp ∗
is proven to be positive definite and radially unbounded in − ω χ(z) − ω̃ω ]
ΦM ΦM
appendix A if kp > 0, ki > 0, β > 0 such that (69) is Rkp
satisfied. +I˜q [−(R + αpq )I˜q − np Ld Id ω ∗ + ω̃
ΦM
Taking advantage from the following cancellations, which Rki
+ χ(z) − Lq I˙q∗ ]
are a direct consequence of the features explained in the ΦM
paragraph after (62): +βχ(z)[np (Ld − Lq )Id I˜q
1
+np (Ld − Lq )Id (−kp ω̃ − ki χ(z)
ΦM
+bω ∗ + τL ) + ΦM I˜q − ki χ(z)]
np LI˜cd I˜cq ω ∗ − np LI˜cq I˜cd ω ∗ = 0,
1
−np C Ṽq Ṽd ω ∗ + np C Ṽd Ṽq ω ∗ = 0, (64) +ω̃[−(b + kp )ω̃ + np Ld Id (−kp ω̃
ΦM
Ṽd np C Ṽq ω̃ − Ṽq − np C Ṽd ω̃ = 0, ˜ ˜
Ṽd Icd − Icd Ṽd = 0, dχ(z) 2
−ki χ(z) + bω ∗ + τL )] + βJ ω̃
−Ṽd I˜d + I˜d Ṽd = 0, Ṽq I˜cq − I˜cq Ṽq = 0, dz
−Ṽq I˜q + I˜q Ṽq = 0, I˜cd np LI˜cq ω̃ − I˜cq np LI˜cd ω̃ = 0, LKpV d ˜ LKpV q ˜
+KiV d ζvd Icd + KiV q ζvq Icq
I˜q [−np Ld Id ω̃ − ΦM ω̃] + Id [np Lq I˜q ω̃ + np Lq Iq∗ ω̃] C C
Lαpq 1
+ω̃[np (Ld − Lq )Id I˜q − np Lq Id Iq∗ + ΦM I˜q ] = 0, +αiq zq + KpV q I˜cq + Lnp KpV d ω I˜cd
Lq Rc
−I˜q αiq zq + αiq zq I˜q = 0, −Id αid zd + αid zd Id = 0, Cαpq
−ω̃ki χ(z) + [ki + β(b + kp )]χ(z)ω̃ + Ṽq
L
q
−βχ(z)(b + kp )ω̃ = 0,
Lαpd 1
+αid zd + KpV d I˜cd − Lnp KpV q ω I˜cq
Ld R
c
Cαpd
+ Ṽd .
Ld
we find that the time derivative of W along the trajectories of
the closed-loop system (56)-(62), (31), (32), (37), (38), (45), Taking into account (30), it is found that Ẇ can be upper
VOLUME 4, 2016 7
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
means that Ẇ ≤ −ν T P ν ≤ 0, if y is small, and the point out that similar problems would arise with the
closed-loop system is stable. Finally, since the closed-loop control design approach introduced by [20] because
system is autonomous, we can invoke the LaSalle invariance such technique also requires to compute and to feedback
principle [24], Ch. 4, to conclude that the origin y = 0 is on line the time derivative of the desired electric currents
locally asymptotically stable. This completes the proof of and voltages.
Proposition 1. • A nested-loop passivity-based control approach is ex-
Conditions for this stability result are summarized by ploited in [19]. This means that the electric current error
kp , ki , αpd , αpq , KpV q , KpV d , Kpcq , Kpcd , Kicq , Kicd , KiV q , is first proven to converge exponentially to zero and this
KiV d , αid , αiq , β, are positive, (69), (54) are satisfied, and allows to use this variable as a vanishing perturbation for
the eight leading principal minors of matrix P introduced in the mechanical subsystem. This, however, requires the
(68) are positive. online computation of the time derivative of the desired
Remark 1: The control scheme in Proposition 1 is made electric current. Instead of that, we use an approach
up of four main loops: 1) a PI controller for electric current which is similar to what was called in [19] passivity-
through the inductor of the DC/DC Buck power converter, based control with total energy shaping. Although the
2) a PI controller for voltage at the DC/DC Buck power latter approach has been disregarded in [19] arguing
converter output (at the capacitor terminals), 3) a PI con- that it results in more complex controllers, we prove the
troller for electric current through the motor stator phase opposite in the present paper.
windings, and 4) a PI controller, with a saturated integral part, • The previous features of our approach allow to natu-
for motor velocity; the saturated integral part is employed rally include PI internal loops, which are important to
in order to render possible to dominate some third order improve the robustness properties of the control scheme
terms, where this integral part appears, with some negative with respect to both parametric uncertainties and exter-
definite second order terms. Thus, our proposal contains nal disturbances.
the fundamental components in industrial applications and,
hence, it is expected to be robust with respect to parametric
uncertainties and external disturbances.
Remark 2: The novel passivity-based approach that is em-
ployed in this paper has the following properties.
V. SIMULATION RESULTS
• The cancellation of terms presented in (64) represents
the energy exchange among the magnetic energy stored
in the motor electrical subsystem, the kinetic energy In order to give some insight on the achievable performance
stored in the motor mechanical subsystem, the electrical by the closed-loop system in Proposition 1, in this sec-
energy stored in the Buck power converter capacitor, tion we present a numerical example. For this, we employ
and the magnetic energy stored in the Buck power the PMSM model Estun EMJ-04APB22, whose numerical
converter inductor. parameters were identified in [25], i.e. np = 4, R =
This property is a direct consequence of the fact that the 2.7[Ohm], Ld = Lq = 8.5[mH], ΦM = 0.301[Nm/A],
first rows of the closed-loop dynamics in (56)-(62) are J = 31.69 × 10−6 [Kgm2 ], b = 52.79 × 10−6 [Nm/(rad/s)].
almost identical to the open-loop dynamics in (9)-(18), For the inverter-Buck DC/DC electronic power converter
excepting (11), (15), (18). Thus, the passivity property system we choose the numerical values employed in [18],
that was established in Section III is instrumental to i.e. L = 4.94[mH], E = 150[V], C = 114.4 × 10−6 [F],
achieve this step. Rc = 48[Ohm]. The controller gains are αpd = 10, αid =
These term cancellations are instrumental to obtain a 30, αpq = 10, αiq = 30, kp = 0.1, ki = 20, KpV q = 40,
simpler control law. This is because, if not cancelled KiV q = 80, KpV d = 40, KiV d = 80, Kpcq = 3000,
naturally, these terms must be cancelled using addi- Kpcd = 3000, Kicq = 2000, Kicd = 2000. Inspired by [23],
tional terms in the control law. Also instrumental for we employ the following linear saturation function:
a simple control law is the fact that our design relies
on dominating many cross terms instead of cancelling
them as usual in differential flatness- and backstepping-
based designs. Moreover, the standard passivity-based
approach introduced by [19] requires to complete iso-
lated error equations for the electrical subsystems by
computing and feeding back online the time derivative
of the desired electrical currents and voltages. As it
is clear in the above proof, this would require a large ∗ ∗ x+L∗
−L + (M − L ) tanh M −L∗ , if x < −L∗
amount of computations which, as it is also remarked
σ(x) = x, if |x| ≤ L∗ ,
by [19], would deteriorate performance because of nu- ∗
merical errors and noise amplification. Finally, let us L∗ + (M − L∗ ) tanh x−L ∗ ,
if x > L∗
M −L
VOLUME 4, 2016 9
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
where M = 43.49, L∗ = 42.26. The desired velocity is ki . We present the transient response in fig. 3 just because the
defined as follows. Given: transient response there can be easily observed.
t − ti1
ti1 = 0.01[s], tf 1 = 0.012[s], a = ,
tf 1 − ti1
450 × 2π
ui1 = 0, uf 1 = [rad/s],
60
ϕ1 = a5 252 − 1050a + 1800a2
−1575a3 + 700a4 − 126a5 ,
t − ti2
ti2 = 0.06[s], tf 2 = 0.063[s], b = ,
tf 2 − ti2
450 × 2π −450 × 2π
ui2 = [rad/s], uf 2 = [rad/s],
60 60 FIGURE 4: Three phase voltages at the converters capacitors
5 2
ϕ2 = b 252 − 1050b + 1800b V1 , V2 , V3 .
−1575b3 + 700b4 − 126b5 ,
In fig. 4 we present the three voltages at the converters ca-
we have that: pacitors. We observe that these three voltages are identical for
0, t < ti1 0 ≤ t < 0.01[s] which clearly shows that these voltages are
ui1 + (uf 1 − ui1 )ϕ1 , ti1 ≤ t < tf 1
not balanced, i.e. that V1 + V2 + V3 6= 0, in this time interval.
ω∗ = uf 1 , tf 1 ≤ t < ti2 . This is because we have assumed that V0 (0) 6= 0 and, since
u i2 + (u f 2 − ui2 )ϕ2 , ti2 ≤ t < tf 2 the zero sequence dynamics is exponentially stable, these
uf 2 , tf 2 ≤ t three voltages become balanced, i.e. V0 → 0 exponentially,
after a while, i.e. after t = 0.01[s] in fig. 4. We also observe
We also consider a step torque disturbance τL = 0.6[Nm]
that any of these three voltages are not larger than ±60[V].
which is applied at t = 0.03[s] and disappears at t = 0.08[s].
We recall that the rated voltage for this motor is ±200[V] rms
Although the electric currents through the motor phase wind-
(see [25]).
ings are always balanced it is interesting to observe the
effects of some unbalanced condition produced by some
disturbance. Moreover, as stated earlier, the source voltages
(at the converters capacitors) and electric currents (through
the converters inductors) may be unbalanced during the tran-
sient response. These are the reasons why all of the initial
conditions are chosen to be zero excepting V0 (0) = 5[V],
I0 (0) = 0.5[A], Ic0 (0) = 1[A]. Finally, we have employed
L0 = 3[mH] < Lq = Ld .
10 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
VI. CONCLUSIONS
We have presented, for the first time, a velocity controller
for a permanent magnet synchronous motor when it is fed by
and inverter-Buck DC/DC power converter system. Although
our stability proof only ensures local asymptotic stability, the
FIGURE 8: Continuous: actual velocity response ω. Dashed: merit of our proposal is that this is the first time that such a
desired velocity ω ∗ . Uncertainty in parameter ΦM is present. problem is solved for an AC motor. Moreover, our controller
is much simpler when compared to proposals in the literature
Finally, in figs. 8, 9, and 10 we present results obtained for DC motors which are designed using differential flatness
when there are uncertainties only in ΦM , only in Rc , and only or backstepping.
VOLUME 4, 2016 11
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
One difficulty to formally solve this problem is that the β > 0 such that:
dq model of the AC motor is required for stability proof Z z
1
purposes whereas formal studies on inverter-Buck DC/DC [ki + β(b + kp )] χ(r)dr − Jβ 2 χ2 (z) ≥
power converter systems are performed in original coordi- τ +bω ∗
− Lk 2
i
nates. Hence, we propose, also for the first time, to perform 1
a dq coordinate transformation on the inverter-Buck DC/DC ≥ [ki + β(b + kp )]G(z) − Jβ 2 H(z) > 0, (69)
2
power converter system in order to succeed to present a
which can be verified graphically (see fig. 11), i.e. that:
stability proof. We believe that this idea may pave the way
to control other classes of AC motors, i.e. induction motors, Z z
1
and we will present such a result elsewhere. [ki + β(b + kp )] χ(r)dr − Jβ 2 χ2 (z),
−
τL +bω ∗
k
2
Another source of difficulty that was found is the fact i
that the three phase source voltages might be unbalanced is a positive definite radially unbounded function in z +
during the transient periods. Recall that the standard model of τL +bω ∗
ki . Thus, since term 12 J(ω̃ + βχ(z))2 ≥ 0 is zero
permanent magnet synchronous motors is derived in the con- only when ω̃ + ∗βχ(z) = 0, property in (69) ensures that ∗
trol literature by assuming that such voltages are balanced. Vω (ω̃, z + τL +bω ) is zero only when both ω̃ and z + τL +bω
ki ki
Hence, we reviewed the permanent magnet synchronous are zero, i.e. when χ(z) = 0. This proves that Vω is positive
motor modeling literature to find how to take into account definite and radially unbounded.
the fact that the three phase source voltages are unbalanced.
.
According to Definition 1:
Lã
|z|, |z| ≤ L∗
|σ(z)| ≥ ,
L∗ , |z| > L∗
and, hence, by direct integration we find that (see fig. 11): (a) σ(z) and χ(z).
Z z
∗
χ(r)dr ≥ G(x),
τL +bω
− ki
'(z
where:
τL +bω ∗ 2
1 z+ , |z| ≤ L∗
2 ki
τ +bω ∗ 2 τ +bω ∗
G(z) = 1 L∗ + L + L∗ + L (z − L∗ ), z > L∗ .
2 ki ki
τ +bω ∗ 2
τ +bω ∗
1 −L∗ + L + −L∗ + L (z + L∗ ), z < −L∗
2 ki ki
Hence, it is clear that it is always possible to find large enough FIGURE 11: Graphical verification of (69).
constants ki > 0 and kp > 0 and a small enough constant
12 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2986701, IEEE Access
VOLUME 4, 2016 13
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.