Standstill Identification of the Rotor Flux in

Salient-Pole PMSMs
Mohamad Koteich Pascal Combes Rashad Ghassani
Schneider Toshiba Inverter Europe Schneider Toshiba Inverter Europe Schneider Toshiba Inverter Europe
Schneider Electric Schneider Electric Schneider Electric
Pacy-sur-Eure, France Pacy-sur-Eure, France Pacy-sur-Eure, France
[email protected] [email protected] [email protected]

Abstract—This paper proposes a new method to estimate the The use of variable speed drives (VSDs) has been steadily
rotor flux of a Permanent-Magnet Synchronous Motor (PMSM) increasing in the past decades, because they ensure high-
at standstill. This method works on motors that have enough performance control, serve as intelligent sensors, and improve
magnetic saliency so that the reluctance torque can cancel the
magnet-alignment torque. A nonlinear controller is designed to the overall efficiency of the motor drive system. Since VSDs
achieve standstill operation. The stability of this controller is are manufactured to work with different types of ac motors,
studied using both the linearized model and nonlinear control parameter identification and self-commissioning are key ca-
theory. It is proven that the desired operating condition, where pabilities to fulfill the requirements of control and monitoring
the rotor flux can be estimated, is locally asymptotically stable. performance [7]. This paper focuses on the identification of the
Thanks to its bifurcation property, the proposed controller
ensures zero speed for any type of PMSM, which is not the case rotor flux. The knowledge of this parameter is important as it
for all existing methods. Furthermore, the estimation accuracy is often used in the feedforward term of the controller, which
is made independent of the motor loading using a compensation contributes to the performance at high speed. Furthermore,
procedure. The theoretical results are verified in simulation on in some applications, this parameter should be monitored to
an 11kW motor model. detect rotor demagnetization.
Index Terms—Permanent-magnet synchronous motor, flux es-
timation, self-commissioning, observability, nonlinear control. State-of-the-art identification methods of the rotor flux in
PMSMs require spinning the rotor [8], [9]. This flux can
then be estimated through the rotational back-emf, which is
I. I NTRODUCTION
proportional to the speed. At standstill, the rotational back-emf
The use of Permanent Magnets (PMs) in Synchronous Mo- is zero, and the flux value cannot be determined. In some ap-
tors (SMs) creates a rotor flux without requiring an excitation plications, the motor can be rotated during the commissioning
current. This results in higher efficiency below the rated speed, phase, nevertheless, in many applications, this is not possible
higher torque and power density, and lower rotor inertia, (cranes, lifts, machines, etc.). Recently, standstill PM flux
compared to other alternating-current (ac) motors [1]. Further- identification has been proposed for salient-pole PMSMs. To
more, the flexibility in the rotor design, and the placement the authors’ knowledge, the first attempt is published in [10]
of the magnets, allow to create a magnetic saliency that can and is studied in this paper. The approach proposed by [11] is
be exploited for torque production [2] and self-sensing of not strictly standstill and may turn the rotor up to a complete
the rotor position, especially in the low-speed range [3], [4]. revolution. Another approach, proposed in [12], is based on
This makes PMSMs the best candidates for compact, high- high-frequency excitation, which we seek to avoid in this
performance applications. paper.
The above strengths come, however, with many challenges This paper presents a formal analysis of the PM flux iden-
that are preventing this motor technology from proliferating tification problem and proposes a new identification method
in the market like induction motors have done [5]; vulnera- based on it. Section II presents the conventional model of the
bility of the rare-earth magnets, risk of demagnetization, and PMSM in the synchronous reference frame, augmented with
high-speed operation issues (degraded efficiency, safety, and the equation of the (constant) PM flux to study its observabil-
damage risks in case of fault). To overcome these weaknesses, ity. The observability study, performed in Section III, suggests
research and development efforts have been made to replace that the permanent-magnet flux is not observable under con-
rare-earth magnets. One viable alternative is the Synchronous ventional cascaded speed/current control structure. Therefore,
Reluctance Motor (SynRM) which can be assisted with non- a new control structure is needed for standstill estimation. This
rare earth magnets to improve its power factor [6]. The scope new controller is proposed and studied in Section IV. It is
of this paper covers both salient-pole Interior-set PMSMs a nonlinear controller, that is subject to bifurcation when the
(IPMSM) and PM assisted SynRMs (PMaSynRM), where the stator current magnitude exceeds a certain value. This property
reluctance torque is comparable to, or greater than, the magnet- turns out to be paramount to guaranteeing the stability of the
alignment torque. desired equilibrium, i.e. the one that allows the identification
of the PM flux. Studies of the Lyapunov function, linearized To ensure LWO, it is sufficient that at least one element
system, and phase portrait are conducted to understand the of the last column of Oy (x) is non-zero, which is the case if
behavior of the controller. The impact of the load torque is ω ̸= 0 or iq ̸= 0. Since we want to remain at standstill, we
studied in Section V and a method to offset it is proposed. must have iq ̸= 0. In a conventional speed control scheme,
The simulation results, presented in Section VI, confirm the this would result in non-zero torque and would tend to rotate
theoretical analysis and the viability of the proposed flux the rotor. Therefore, a new controller needs to be designed for
estimator. that purpose, i.e. to change the operating point at standstill
from iq = 0 to iq ̸= 0 without generating torque.
II. PMSM M ODEL
The motor is modeled under the classical assumptions of a IV. C ONTROLLER D ESIGN AND DYNAMICS
linear magnetic circuit, negligible iron losses, and sinusoidal The PMSM electromagnetic torque equation is [15]:
back-EMF. Furthermore, we assume, for rotor flux identifica- 3 3
tion purposes, that: Tem = np (L∆ id + ϕr )iq = np ψda iq (2)
2 2
• The rotor position and speed are measured accurately;
where ψda = (L∆ id + ϕr ) denotes the active flux, i.e. the
• The mechanical load and losses are negligible.
torque-producing flux. The motivating idea behind this new
The PMSM model in the rotor flux reference frame is [13]: speed controller is to ensure ψda = 0 so that we can have
did Tem = 0 with iq ̸= 0. Once this desired operating point is
Ld = vd − Rs id + ωLq iq (1a)
dt achieved, we can obtain ϕr assuming that we know Ld and
diq Lq . Their identification is beyond the scope of this paper, but
Lq = vq − Rs iq − ω(Ld id + ϕr ) (1b)
dt note that it is possible to identify Ld and Lq at this operating
dω 3 point, once the controller is at steady-state, by signal injection
J = n2p (ϕr + L∆ id )iq (1c)
dt 2 techniques. It is worth mentioning that the whole inductance
dϕr and/or flux maps are not necessary.
=0 (1d)
dt In PMSMs with high saliency (PMaSynRMs and IPMSMs),
with L∆ = Ld − Lq . This model is a nonlinear state-space |L∆ id | is comparable to, or greater than, |ϕr | and the active
system, with the state vector x = [id , iq , ω, ϕr ]T , the input flux can be zeroed. This is not possible for non-salient PMSMs
vector u = [vd , vq ]T , and the (linear) output y = [id , iq , ω]T . (L∆ = 0), for which the active flux is constant, and impractical
ω is the measured electric angular frequency. The goal is to for low-saliency PMSMs since the current magnitude should
identify (estimate) the rotor PM flux ϕr using the available remain smaller than Ismax . In the sequel, we define the current
output y. Id∗ as:
ϕr
III. O BSERVABILITY A NALYSIS Ismax ≥ Id∗ = ≥0
|L∆ |
First, we need to study the observability of the system.
A. Controller Design in the Cartesian Coordinates
Indeed, Observability conditions are the mathematical formu-
lation of the capability of estimating the system state x using One can use the conventional cascaded speed controller,
the available measurement y and input u. The system (1) but instead of using i∗q as the main control variable for the
is nonlinear, therefore, we can apply the rank criterion [14] torque, i∗d is used, as shown in Fig. 1. This controller was
to calculate the sufficient local weak observability (LWO) proposed by [10], but not thoroughly analyzed. The gain Kd =
3 ∗
conditions. Evaluating the observability matrix Oy (x) up to 2 np L∆ iq is applied to transform the speed controller output
the second-order derivative yields (torque) into the current i∗d .

1 0 0 0
 
0 1 0 0 Kp
ω∗ = 0

0 0 1 1 i∗d vd id
0 +
 

+ + Kd
∗ ∗ ∗ 0
 
 − Ki Current iq
ω PMSM

∗ ∗ ∗
 ω s i∗q Control
vq
−  ω

 Lq 

 3 i
2 q

Oy =  ∗ ∗ ∗ n p

 2 2J 
3 2 Lq iq ω2
 
∗ ∗ ∗ −
 
np  Fig. 1. Controller in Cartesian coordinates [10]
 2 L d J Ld 
∗ ∗ ∗ Rs ω − 3 n2 iq (2ϕ − (2L − L )i )
 
r d q d
The shortcomings of this method are:
L2q 2 p Lq J
 
max
  • If the current Id∗ is not reachable, i.e. Is < Id∗ , the
 3 2 1 
∗ ∗ ∗ − np ω(L∆ id + ϕr ) motor will start rotating in an uncontrolled way, as the
2 Lq J system does not have any equilibrium in this case. To
where the asterisk ∗ replaces the elements that are less relevant avoid this situation on low saliency motors, this method
for the calculation of the rank of Oy (x). requires the knowledge of Id∗ , which depends on ϕr .
 • Further measures should be taken to avoid exceeding the λ1 λ2 > 0, where λ1 and λ2 are the eigenvalues of A. The
current limit. trace and determinant expressions are1 :
We would rather have a new controller which, somehow,
tr A = −K2 ϕr sin γ0 + L∆ Is − 2L∆ Is sin2 γ0


does not apply any torque until the point Id∗ is reached (by
det A = K1 K3 ϕr sin γ0 + L∆ Is − 2L∆ Is sin2 γ0


setting i∗q = 0) and then setting id and iq to the right values
when Ismax > Id∗ . Such a bifurcation can only happen with a
nonlinear controller. Note that the relation between the determinant and the trace
is independent of the saliency:
B. Controller Design in Polar Coordinates
K1 K3 Ki
The proposed method is based on the torque equation in det A = − tr A = − tr A (6)
K2 Kp
polar coordinates (Is , γ):
3 This relation can be plotted as a line with negative slope,
Tem = np (ϕr − L∆ Is sin γ)Is cos γ going through the origin, on the trace-determinant diagram
2
as shown in Fig. 3. The impact of the speed loop tuning (Kp
with and Ki ), when det A > 0, is to define whether the stable
id = −Is sin γ ; iq = Is cos γ equilibrium is a nodal sink or a spiral sink. For det A < 0, the
The current angle γ is used as the control variable, whereas equilibirum is a saddle point (unstable). The types of equilibria
Is is set to a predefined value. This is illustrated in Fig. 2. are summarized in Table I.
Note that the gain Kγ = − 32 np L∆ Is is applied to match the
TABLE I
speed controller output with the current phase angle. T YPES OF EQUILIBRIA . F OR SADDLE POINTS det A < 0 AND tr A > 0.
1) System state-space model: The current loop is faster than F OR STABLE POINTS det A > 0 AND tr A < 0
the speed loop in a cascaded system, so we can assume that
Is < Id∗
id = i∗d and iq = i∗q . The system/controller model is: Equilibrium
L∆ > 0 L∆ < 0
Is > Id∗
x0,1 = (π/2, 0) stable saddle saddle
dγ Ki Kp 3n2p x0,2 = (−π/2, 0) saddle stable saddle
=− ω− Is (ϕr − L∆ Is sin γ) cos γ
dt Kγ Kγ 2J x0,3 N/A N/A stable
x0,4 N/A N/A stable
dω 3n2p
= Is (ϕr − L∆ Is sin γ) cos γ
dt 2J
To simplify, the above system can be written as follows: The system is subject to bifurcation that is very useful
for the standstill identification of the rotor flux. The stable
dγ
= K1 ω + K2 (ϕr − L∆ Is sin γ) cos γ (3a) equilibria are shown on the current circle diagram of Fig. 4.
dt We can see that before reaching Id∗ , the stable equilibrium
dω
= K3 (ϕr − L∆ Is sin γ) cos γ (3b) consists of iq = 0, and for Is > Id∗ the equilibrium gives
dt
|id | = Id∗ .
where K1 and K2 have the same sign as L∆ , and K3 > 0. 4) Lyapunov stability: Consider the following function
Note that K2 = −(Kp /Kγ )K3 and K1 = (Ki /Kp )K2 /K3 .
2) Equilibria: The above system has four equilibrium 1 2 1 K1 2 1
V (ω, γ) = (ϕr − L∆ Is sin γ) + Is L∆ ω − V0 ≥ 0
points x0 = (γ0 , ω0 ): 2 2 K3 2
(7)
x0,1 = (π/2, 0) with the constant V0 = max(0, ϕr − |L∆ |Is )2 . Computing its
x0,2 = (−π/2, 0) derivative yields
 
ϕr 2
x0,3 = sin−1 ,0 V̇ (ω, γ) = −Is L∆ K2 ((ϕr − L∆ Is sin γ) cos γ) ≤ 0 (8)
L∆ Is
 
−1 ϕr When Is < Id∗ and L∆ > 0, in a neighborhood around x0,1 ,
x0,4 = π − sin ,0
L∆ Is we have V (x ̸= x0,1 ) > 0, V (x0,1 ) = 0, V̇ (x ̸= x0,1 ) < 0
and V̇ (x0,1 ) = 0. Consequently [16, theorem 4.1] shows that
The last two equilibria, x3 and x4 , only exist if Is > Id∗ .
x0,1 is asymptotically stable. When Is < Id∗ and L∆ < 0,
3) Local Stability of Equilibrium: The linearization of the
mutatis mutandis, it can be shown that x0,2 is asymptotically
system about (γ0 , ω0 ) gives:
stable. Finally, when Is > Id∗ , applying [16, theorem 4.1] on
neighborhoods around x0,3 or x0,4 , where we have V (x ̸=
    
d γ −K2 (ϕr sin γ0 + L∆ Is cos 2γ0 ) K1 γ
= (4)
dt ω −K3 (ϕr sin γ0 + L∆ Is cos 2γ0 ) 0 ω x0,i ) > 0, V (x0,i ) = 0, V̇ (x ̸= x0,i ) < 0 and V̇ (x0,i ) = 0,
| {z } with i ∈ {3, 4}, proves that x0,3 and x0,4 are asymptotically
A
stable.
The stability conditions can be calculated using the trace-
determinant criterion, i.e. tr A = λ1 + λ2 < 0 and det A = 1 cos 2γ
0 = 1 − 2 sin2 γ0
 Is i∗d vd id
Kp sin()
ω∗ = 0 1 Current iq
+ Control PMSM
+ + Kγ i∗q vq ω
− Ki cos()
ω s

Fig. 2. Proposed controller for standstill flux estimation.

We assume that fl (Ω = 0) = 0 and its derivative fl′ (Ω) > 0.

 
Bf
D = −
Ki
Kp
T +
J
det A
with load When linearized about Ω0 = 0, the equation becomes:
Stable Spiral Untable Spiral 
J dω 1 ∂fl 
Center

D = T 2 /4
= Tem − ω − Tl0
np dt np ∂Ω Ω=0
| {z }
Bf

A. Speed-dependent load
Stable Node Unstable Node First, we consider only the impact of speed-dependent load
Bf
(such as viscous friction and aerodynamic friction), i.e. Tl0 =
−
J
tr A 0. The system can be written as:
K
D = − i T
Kp
without load dγ K 2 np
= K2 (ϕr − L∆ Is sin γ) cos γ + K1 ω − fl (ω/np )
Unstable, Saddle Point dt K3 J
dω np
Fig. 3. The trace-determinant plane. D = det A and T = tr A = K3 (ϕr − L∆ Is sin γ) cos γ − fl (ω/np )
dt J
Notice that the speed-dependent load torque does not change
iq
the equilibrium states as long as the integral gain in K1 is not
zero. After linearization, the new expressions of the trace and
Ismax determinant are:
Is2
tr A = −K4 − K2 ϕr sin γ0 + L∆ Is − 2L∆ Is sin2 γ0


Tl0
det A = K1 K3 ϕr sin γ0 + L∆ Is − 2L∆ Is sin2 γ0


Is1
id with K4 = Bf /J. The relation between the trace and the
Id∗
determinant becomes:
0  
K1 K3 Ki Bf
det A = − (tr A + K4 ) = − tr A + (11)
K2 Kp J
This line is plotted in blue in Fig. 3. The friction will
not change the slope, but will shift the line to the left. So
oscillating spirals tend to be nodes under load. This does not
impact the stability of the equilibrium (since det A > 0).

Fig. 4. Equilibrium points as Is increases in the case where Ld < Lq .

B. Speed-independent load
The blue lines correspond to the system without speed-independent load, we If we consider Tl0 ̸= 0, the stability of the system will not
notice the bifurcation when Is achieves Id∗ . Note that the plot for Ld > Lq
is obtained by mirroring the above plot with respect to the q-axis.
be impacted, since the trace and the determinant remain the
same, and the rotor will remain at standstill. But the current
angle at equilibrium will be modified and the identified rotor
V. T HE I MPACT OF L OAD T ORQUE flux ϕ̂r will be incorrect. At equilibrium, we have ω0 = 0 and:
If we consider a general load torque, composed of a speed- 3 Tl0
np (ϕr + L∆ id )iq = Tl0 ⇒ ϕr = 3 − L∆ id
dependent torque fl (Ω) and a speed-independent torque Tl0 , 2 2 np iq
the equation of motion becomes: The above equation is a fourth-degree equation. Analytically
J dω computing its equilibrium points is tedious. Nevertheless, we
= Tem − fl (ω/np ) − Tl0 can picture them graphically (Fig. 4). Equilibrium exists only
np dt
if the current Is is high enough to produce a torque Tem ≥
Tl0 , otherwise the speed is not zero. Above this threshold, the
equilibrium points are the intersection of the torque curve with
the current circle. Up to a certain value of Is , the current circle
intersects only twice with the torque curve (e.g. Is1 in Fig. 4).
Beyond that value, four intersections occur, corresponding to
four equilibrium points (e.g. Is2 in Fig. 4). By simulation (see
Section VI), we found that the points near the line id = Id∗
are stable equilibria, but the formal proof of this fact is beyond
the scope of this paper. In the sequel, we will focus on the
compensation of the flux estimation error due to Tl0 .
If we estimate the flux as ϕ̂r0 = −L∆ id , the estimation
error is: Fig. 5. Relative rotor flux estimation error as a function of the load torque
2 Tl0
εϕr = ϕr − ϕ̂r0 = (12) Tl0 and the current norm Is .
3 np iq
This error, computed numerically using the motor parameters
in Table II, is shown in Fig. 5. Unsurprisingly, the higher on Fig.7, and using equation (14), the estimated rotor flux is
the load torque Tl0 , the higher the error, and the error can ϕ̂r = 0.337W b, which is more accurate than the estimates
be reduced by increasing the current norm Is . This error can (−L∆ id1 ) and (−L∆ id2 ).
be cancelled by applying the method on two different current The evolution of the rotor position during the identification
levels. Indeed, we obtain procedure is shown in Fig. 8; the transient of the first step Is1
3 3 generated a torque that moved the rotor of less than 0.01rad,
np (ϕr + L∆ id1 )iq1 = Tl0 = np (ϕr + L∆ id2 )iq2 (13) and the position remained almost constant till the end of the
2 2
procedure. This rotor displacement is negligible and standstill
and solving for ϕr condition is respected.
id1 iq1 − id2 iq2 Note that for this test, both the Cartesian-coordinates-based
ϕ̂r = −L∆ (14)
iq1 − iq2 method, and the nonlinear polar-coordinates-based method
Ideally, the two current points are as different as possible, so exhibit a similar behavior and yield similar results.
that the denominator of this formula is maximized. Tl0 can be
estimated by solving for it in (12) and (13): TABLE II
11 KW PM A -S YN RM PARAMETERS
3 id1 − id2
T̂l0 = − np L∆ iq1 iq2 Parameter Symbol Value Unit
2 iq1 − iq2 Nominal Power Pnom 11 kW
Notice the parameter L∆ may change between the two Rated Voltage Vnom 400 V
Rated Current Inom 21 A
operating points, due to magnetic saturation. In this case, the Rated Torque Tnom 70 Nm
equation (14) shall be replaced by the more general one Rated Speed Nnom 1500 RPM
L∆1 id1 iq1 − L∆2 id2 iq2 Pole Pairs np 2 -
ϕ̂r = − (15) Stator Resistance Rs 0.309 Ω
iq1 − iq2 D-axis Inductance Ld 7.63 mH
Q-axis Inductance Lq 47 mH
Even in this case the load torque can be estimated using Rotor Flux ϕr 0.34 Wb
3 L∆1 id1 − L∆2 id2 Rotor Inertia J 0.055 kg.m2
T̂l0 = − np iq1 iq2 Friction Coefficient Bf 0.005 Nm.s
2 iq1 − iq2

VI. S IMULATION R ESULTS

The proposed method was tested in simulation, using Mat- VII. C ONCLUSIONS
lab/Simulink, on the 11kW-motor model whose parameters are The analysis carried out in this paper shows that the
presented in II. A speed-independent load torque of 1%Tnom proposed standstill PM flux identification method is viable.
is applied. The nonlinear controller described in Fig. 2 is Although it is limited to PMSMs that are salient enough,
used, and the current magnitude Is∗ is changed step-wise two so that the current needed to nullify the active flux is less
∗ than the maximum current, it can be applied safely to other
times, the first, Is1 = 1.5Id∗ , at the starting time, t = 0sec.
∗ types of synchronous motors as it guarantees both the standstill
(Fig. 6), and the second, Is2 = 2Id∗ , at t = 1sec. (Fig. 7).
Note that the choice of Is1 and Is2 is for illustration purpose. condition and the control of the current magnitude.
In practice Id∗ is unknown and a ramp, or consecutive steps, The accuracy of the flux estimation depends on the accuracy
on Is can be applied. of the estimated parameters Ld and Lq , which are subject
Based on the dq-currents values, (id1 = 8.072A, iq1 = to magnetic saturation. This can be overcome without the
10.117A) shown in Fig. 6 and (id2 = 8.238A, iq2 = 15.165A) need for inductance/flux maps by applying an identification
 algorithm using voltage signal injection whenever the desired
operating point is reached. This may be addressed in the future.
Finally, this study is based on the assumption of accurate
position and speed measurement. A sensorless version is
needed to cover a wider range of industrial applications. Since
the addressed motors have high saliency, we expect high-
frequency signal injection techniques to provide satisfactory
results.
R EFERENCES
[1] Melfi, Michael J., Steve Evon, and Robbie McElveen. ”Induction versus
permanent magnet motors.” IEEE Industry Applications Magazine 15.6
(2009): 28-35.
[2] Jahns, Thomas M., Gerald B. Kliman, and Thomas W. Neumann.
”Interior permanent-magnet synchronous motors for adjustable-speed
drives.” IEEE Transactions on Industry Applications 4 (1986): 738-747.
[3] Corley, Matthew J., and Robert D. Lorenz. ”Rotor position and velocity
estimation for a salient-pole permanent magnet synchronous machine at
standstill and high speeds.” IEEE transactions on industry applications
34.4 (1998): 784-789.
[4] F. Briz and M. W. Degner, ”Rotor Position Estimation,” in IEEE
Industrial Electronics Magazine, vol. 5, no. 2, pp. 24-36, June 2011.
[5] Jahns, Thomas. ”Getting rare-earth magnets out of EV traction ma-
chines: A review of the many approaches being pursued to minimize
Fig. 6. DQ-currents during the first step Is1 . or eliminate rare-earth magnets from future EV drivetrains.” IEEE
Electrification Magazine 5.1 (2017): 6-18.
[6] Murataliyev, Mukhammed, et al. ”Synchronous Reluctance Machines:
A Comprehensive Review and Technology Comparison.” Proceedings
of the IEEE (2022).
[7] S. A. Odhano, P. Pescetto, H. A. A. Awan, M. Hinkkanen, G. Pellegrino
and R. Bojoi, ”Parameter Identification and Self-Commissioning in AC
Motor Drives: A Technology Status Review,” in IEEE Transactions on
Power Electronics, vol. 34, no. 4, pp. 3603-3614, April 2019.
[8] Z. Q. Zhu, X. Zhu, P. D. Sun and D. Howe, ”Estimation of Winding Re-
sistance and PM Flux-Linkage in Brushless AC Machines by Reduced-
Order Extended Kalman Filter,” 2007 IEEE International Conference on
Networking, Sensing and Control, 2007, pp. 740-745.
[9] S. -M. Yang and K. -W. Lin, ”Automatic Control Loop Tuning for
Permanent-Magnet AC Servo Motor Drives,” in IEEE Transactions on
Industrial Electronics, vol. 63, no. 3, pp. 1499-1506, March 2016.
[10] S. A. Odhano, P. Giangrande, R. I. Bojoi and C. Gerada, ”Self-
Commissioning of Interior Permanent- Magnet Synchronous Motor
Drives With High-Frequency Current Injection,” in IEEE Transactions
on Industry Applications, vol. 50, no. 5, pp. 3295-3303, Sept.-Oct. 2014.
[11] J. Jacob, P. Kumar, S. Calligaro and R. Petrella, ”Self-Commissioning
Identification of Permanent Magnet Flux-Linkage Magnitude in Sensor-
less Drives for PMSM at Quasi Stand-Still,” 2018 IEEE 9th International
Symposium on Sensorless Control for Electrical Drives (SLED), 2018,
pp. 144-149.
[12] P. Pescetto and G. Pellegrino, ”Determination of PM Flux Linkage Based
on Minimum Saliency Tracking for PM-SyR Machines Without Rotor
Movement,” in IEEE Transactions on Industry Applications, vol. 56, no.
5, pp. 4924-4933, Sept.-Oct. 2020.
[13] Pillay, Pragasen, and Ramu Krishnan. ”Modeling, simulation, and
Fig. 7. DQ-currents during the second step Is2 . analysis of permanent-magnet motor drives. I. The permanent-magnet
synchronous motor drive.” IEEE Transactions on industry applications
25.2 (1989): 265-273.
[14] Hermann, R. and Krener, A. J. (1977). Nonlinear controllability and
observability. Automatic Control, IEEE Transactions on, 22(5):728–740.
[15] Boldea, Ion, Mihaela Codruta Paicu, and Gheorghe-Daniel Andreescu.
”Active flux concept for motion-sensorless unified AC drives.” IEEE
Transactions on Power Electronics 23.5 (2008): 2612-2618.
[16] Khalil, Hassan K. ”Nonlinear Systems”, 3rd edition, Pearson Education.
2002.

Fig. 8. Rotor’s mechanical position during the identification procedure.